<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
    <title>cam@arch</title>
    <subtitle>Notes from a homelab, and whatever else.</subtitle>
    <link rel="self" type="application/atom+xml" href="https://legitbiz.xyz/atom.xml"/>
    <link rel="alternate" type="text/html" href="https://legitbiz.xyz"/>
    <generator uri="https://www.getzola.org/">Zola</generator>
    <updated>2026-10-04T00:00:00+00:00</updated>
    <id>https://legitbiz.xyz/atom.xml</id>
    <entry xml:lang="en">
        <title>Getting Started with bettercap: A Home-Lab Tutorial</title>
        <published>2026-10-04T00:00:00+00:00</published>
        <updated>2026-10-04T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/bettercap-home-lab/"/>
        <id>https://legitbiz.xyz/posts/bettercap-home-lab/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/bettercap-home-lab/">&lt;p&gt;&lt;a rel=&quot;external&quot; href=&quot;https://www.bettercap.org/&quot;&gt;bettercap&lt;/a&gt; is an open-source network reconnaissance
and security-testing framework written in Go. It bundles tools for scanning and
attacking Ethernet LANs, Wi-Fi networks, Bluetooth Low Energy devices, and
wireless HID devices into one interactive console. This tutorial walks you
through setting up a safe lab and using bettercap&#39;s core features: wired recon, a
man-in-the-middle attack, Wi-Fi scanning, and WPA handshake capture. Every
exercise runs against hardware and networks you own, inside your own home lab.
And to keep the workflow fast, we&#39;ll drive bettercap through
&lt;a rel=&quot;external&quot; href=&quot;https://github.com/NPCmillionaire/bettercaptui&quot;&gt;bettercaptui&lt;/a&gt;, a terminal
frontend that puts bettercap&#39;s modules behind a live, keyboard-driven interface.&lt;/p&gt;
&lt;h2 id=&quot;before-you-start-stay-legal&quot;&gt;Before you start: stay legal&lt;/h2&gt;
&lt;p&gt;bettercap&#39;s recon features are harmless, but its attack features (ARP spoofing,
DNS spoofing, deauthentication, handshake capture) are only legal against
networks and devices you own or have explicit written permission to test.
Running them against someone else&#39;s network is a crime in most jurisdictions,
including under the U.S. Computer Fraud and Abuse Act, regardless of intent.&lt;/p&gt;
&lt;p&gt;This tutorial keeps everything inside a self-contained lab:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The wired exercises target a VM you created on your own machine.&lt;/li&gt;
&lt;li&gt;The Wi-Fi exercises target a spare access point you set up yourself (an old
router or a phone hotspot works).&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Don&#39;t point any attack module at your household&#39;s main network or any device you
don&#39;t control. A deauth, for example, will knock real devices offline.&lt;/p&gt;
&lt;h2 id=&quot;how-bettercap-is-organized&quot;&gt;How bettercap is organized&lt;/h2&gt;
&lt;p&gt;bettercap is a single console that loads &lt;strong&gt;modules&lt;/strong&gt;, each owning a namespace of
commands. You turn a module on, configure it with &lt;code&gt;set&lt;/code&gt;, and read its output.
The ones this tutorial uses:&lt;/p&gt;
&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Module&lt;/th&gt;&lt;th&gt;Namespace&lt;/th&gt;&lt;th&gt;What it does&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;
&lt;tr&gt;&lt;td&gt;Net recon&lt;/td&gt;&lt;td&gt;&lt;code&gt;net.*&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Discover and probe hosts on the LAN&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td&gt;Sniffer&lt;/td&gt;&lt;td&gt;&lt;code&gt;net.sniff&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Capture and parse packets&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td&gt;ARP spoofer&lt;/td&gt;&lt;td&gt;&lt;code&gt;arp.spoof&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Redirect a target&#39;s traffic through you&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td&gt;DNS spoofer&lt;/td&gt;&lt;td&gt;&lt;code&gt;dns.spoof&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Answer DNS queries with addresses you choose&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td&gt;Wi-Fi&lt;/td&gt;&lt;td&gt;&lt;code&gt;wifi.*&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Scan APs and clients, deauth, capture handshakes&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td&gt;Web UI&lt;/td&gt;&lt;td&gt;&lt;code&gt;api.rest&lt;/code&gt;, &lt;code&gt;http.server&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Serve the browser dashboard&lt;/td&gt;&lt;/tr&gt;
&lt;/tbody&gt;&lt;/table&gt;
&lt;p&gt;Two conventions make the console quick to use. Commands separated by &lt;code&gt;;&lt;/code&gt; run in
sequence, so you can configure and launch in one line. And a &lt;strong&gt;caplet&lt;/strong&gt; is just a
&lt;code&gt;.cap&lt;/code&gt; file of these commands that bettercap replays, which is how you save a
setup you use often. Tab completion works for every command.&lt;/p&gt;
&lt;h2 id=&quot;driving-bettercap-with-bettercaptui&quot;&gt;Driving bettercap with bettercaptui&lt;/h2&gt;
&lt;p&gt;The raw console is powerful but spartan: you type a command, read a wall of text,
type another. &lt;a rel=&quot;external&quot; href=&quot;https://github.com/NPCmillionaire/bettercaptui&quot;&gt;bettercaptui&lt;/a&gt;
wraps that same session in a terminal UI, so the things you&#39;d otherwise juggle by
hand happen in panels that update on their own:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Live host and AP tables that refresh as recon runs, instead of re-typing
&lt;code&gt;net.show&lt;/code&gt; or &lt;code&gt;wifi.show&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Modules you toggle with a keystroke rather than &lt;code&gt;module on&lt;/code&gt; / &lt;code&gt;module off&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;The event and capture stream in its own pane, so a handshake or a sniffed
credential doesn&#39;t scroll away.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;It talks to bettercap over the same REST API the web UI uses, so everything in
this tutorial works identically whether you type the commands or trigger them
from the TUI. Install it from the repo:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;git&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; clone https://github.com/NPCmillionaire/bettercaptui&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #F38BA8;font-style: italic;&quot;&gt;cd&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; bettercaptui&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #9399B2;font-style: italic;&quot;&gt;# follow the README to build and run&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;For each exercise below, the console commands are shown so you understand what&#39;s
happening underneath; in bettercaptui you&#39;ll find the same actions on screen.
Start bettercap&#39;s API, then point the TUI at it.&lt;/p&gt;
&lt;h2 id=&quot;setting-up-the-lab&quot;&gt;Setting up the lab&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Your machine.&lt;/strong&gt; Kali or Parrot OS ship bettercap and the supporting tools, so
they&#39;re the easiest starting point. On Arch it&#39;s a single package (&lt;code&gt;sudo pacman -S bettercap&lt;/code&gt;). On Debian/Ubuntu: &lt;code&gt;sudo apt install bettercap&lt;/code&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;A target for the wired exercises.&lt;/strong&gt; Create one VM (any Linux is fine) on the
same host-only or NAT network as your bettercap machine. This is the &quot;victim&quot;
you&#39;ll safely spoof. Note its IP.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Hardware for the Wi-Fi exercises.&lt;/strong&gt; You need a wireless adapter that supports
&lt;strong&gt;monitor mode&lt;/strong&gt; and packet injection. The built-in card on most laptops does
not. Common lab choices use the Atheros AR9271 or Ralink RT3070 chipsets. Confirm
your interface name with &lt;code&gt;iw dev&lt;/code&gt; (often &lt;code&gt;wlan0&lt;/code&gt;).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;A target AP for the Wi-Fi exercises.&lt;/strong&gt; Set up a spare router or a phone hotspot
with a WPA2 password you choose. This is the only Wi-Fi network you&#39;ll attack.&lt;/p&gt;
&lt;p&gt;Verify the install before moving on:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;sudo&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; bettercap -version&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;h2 id=&quot;starting-the-console&quot;&gt;Starting the console&lt;/h2&gt;
&lt;p&gt;Launch bettercap on your LAN interface (use &lt;code&gt;ip addr&lt;/code&gt; to find it, often &lt;code&gt;eth0&lt;/code&gt; or
&lt;code&gt;enp0s3&lt;/code&gt;):&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;sudo&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; bettercap -iface eth0&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;You land at an interactive prompt showing your network and IP. A few orientation
commands:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;help              # list all modules and their on/off state&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;help net.recon    # show one module&amp;#39;s commands and options&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;active            # show which modules are currently running&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;To configure a module you &lt;code&gt;set&lt;/code&gt; one of its parameters; to see a parameter&#39;s
current value, type its name alone. Type &lt;code&gt;quit&lt;/code&gt; to exit, which also cleanly stops
any running module (important, since a spoofer leaves the target&#39;s network
settings changed until it restores them on exit).&lt;/p&gt;
&lt;h2 id=&quot;part-1-recon-on-your-own-lan&quot;&gt;Part 1: Recon on your own LAN&lt;/h2&gt;
&lt;p&gt;Start by mapping the network. &lt;code&gt;net.probe&lt;/code&gt; sends probe packets to every address in
your subnet; &lt;code&gt;net.show&lt;/code&gt; prints what answered.&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;net.probe on&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;net.show&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;code&gt;net.show&lt;/code&gt; lists each host&#39;s IP, MAC, hardware vendor, and the traffic bettercap
has seen. Find your target VM in the list.&lt;/p&gt;
&lt;p&gt;Now watch live traffic with the sniffer. Left wide open it captures everything on
your segment, so scope it to your VM:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set net.sniff.filter &amp;quot;host 192.168.56.20&amp;quot;&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;net.sniff on&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;The &lt;code&gt;filter&lt;/code&gt; takes standard BPF syntax (the same as tcpdump). bettercap parses
common protocols as they pass and prints a readable line per packet: DNS lookups,
HTTP requests, and any plaintext credentials crossing the wire. Generate some
traffic on the VM and watch it appear. Turn it off with &lt;code&gt;net.sniff off&lt;/code&gt;.&lt;/p&gt;
&lt;h2 id=&quot;part-2-a-man-in-the-middle-attack-on-your-vm&quot;&gt;Part 2: A man-in-the-middle attack on your VM&lt;/h2&gt;
&lt;p&gt;ARP spoofing tells your target VM that &lt;em&gt;you&lt;/em&gt; are the router, so its traffic flows
through your machine first. This is the classic MITM, and here it runs only
against the VM you own.&lt;/p&gt;
&lt;p&gt;First enable IP forwarding so the VM keeps working while you&#39;re in the middle
(otherwise you black-hole its traffic):&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;sudo&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; sysctl -w net.ipv4.ip_forward=&lt;/span&gt;&lt;span style=&quot;color: #FAB387;&quot;&gt;1&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Then, in bettercap, point the spoofer at the VM and turn it on:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set arp.spoof.targets 192.168.56.20&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;arp.spoof on&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;net.sniff on&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;With traffic now passing through you, the sniffer sees the VM&#39;s unencrypted
requests.&lt;/p&gt;
&lt;p&gt;To show why DNS matters, add DNS spoofing. This answers the VM&#39;s lookups for a
chosen domain with your own IP:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set dns.spoof.domains example.com&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set dns.spoof.address 192.168.56.10&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;dns.spoof on&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Browse to &lt;code&gt;example.com&lt;/code&gt; on the VM and it resolves to your machine. When you&#39;re
done, stop cleanly so bettercap restores the VM&#39;s real ARP entries:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;arp.spoof off&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;dns.spoof off&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;strong&gt;The lesson:&lt;/strong&gt; this is why HTTPS and DNSSEC exist. Modern sites use HSTS, so the
browser refuses the downgrade, which is the defense working as designed.&lt;/p&gt;
&lt;h2 id=&quot;part-3-wi-fi-recon&quot;&gt;Part 3: Wi-Fi recon&lt;/h2&gt;
&lt;p&gt;Switch to your monitor-mode adapter. bettercap will put the interface into
monitor mode for you when you start it on a wireless interface:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;sudo&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; bettercap -iface wlan0&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Start scanning and show what&#39;s around:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wifi.recon on&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wifi.show&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;code&gt;wifi.show&lt;/code&gt; lists access points with their BSSID, SSID, channel, encryption, and
signal strength, plus the clients seen talking to each one. Because scanning hops
across channels, you&#39;ll see the list grow over a few seconds.&lt;/p&gt;
&lt;p&gt;A couple of controls make the output manageable:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set wifi.show.sort rssi asc      # strongest signal last&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wifi.recon.channel 1,6,11        # lock to common 2.4GHz channels&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Find your own test AP in the list and note its &lt;strong&gt;BSSID&lt;/strong&gt; and &lt;strong&gt;channel&lt;/strong&gt;. That&#39;s
your only target for the next part.&lt;/p&gt;
&lt;h2 id=&quot;part-4-capturing-a-wpa-handshake-from-your-own-ap&quot;&gt;Part 4: Capturing a WPA handshake from your own AP&lt;/h2&gt;
&lt;p&gt;A WPA/WPA2 handshake is the four-message exchange a client and AP perform when
the client joins. Capturing it lets you later test your own password&#39;s strength
offline. bettercap writes captures to &lt;code&gt;~/bettercap-wifi-handshakes.pcap&lt;/code&gt; by
default.&lt;/p&gt;
&lt;p&gt;Lock onto your test AP&#39;s channel so you don&#39;t miss frames while hopping:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wifi.recon.channel 6&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;strong&gt;Option A, no client needed (PMKID).&lt;/strong&gt; Many APs leak a PMKID that yields the
same crackable material without any client. Associate with your AP and bettercap
grabs it:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wifi.assoc AA:BB:CC:DD:EE:FF&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;strong&gt;Option B, deauth to force a reconnect.&lt;/strong&gt; If a device of yours is connected,
briefly deauthenticate it; when it automatically rejoins, you capture the
handshake:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wifi.deauth AA:BB:CC:DD:EE:FF&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Use &lt;code&gt;AA:BB:CC:DD:EE:FF&lt;/code&gt; as your AP&#39;s real BSSID. Only deauth your own AP and your
own client, this is the one command most likely to disrupt real devices if aimed
wrong.&lt;/p&gt;
&lt;p&gt;bettercap prints a line when it captures a handshake or PMKID. From there you&#39;d
run the &lt;code&gt;.pcap&lt;/code&gt; through a cracker like hashcat against a wordlist, purely to
confirm your own passphrase isn&#39;t weak. (Cracking is a separate topic and outside
this tutorial.)&lt;/p&gt;
&lt;h2 id=&quot;caplets-the-ticker-and-the-web-ui&quot;&gt;Caplets, the ticker, and the web UI&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Caplets&lt;/strong&gt; save a workflow. List and update the bundled ones:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;caplets.show&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;caplets.update&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Write your own by putting commands in a &lt;code&gt;.cap&lt;/code&gt; file and loading it at launch. A
lab recon caplet, &lt;code&gt;lanscan.cap&lt;/code&gt;:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;net.probe on&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set net.sniff.filter &amp;quot;not arp&amp;quot;&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;net.sniff on&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;sudo&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; bettercap -iface eth0 -caplet lanscan.cap&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;strong&gt;The ticker&lt;/strong&gt; re-runs commands on a timer, handy for a live view:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;set ticker.commands &amp;quot;clear; wifi.show&amp;quot;&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;ticker on&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;strong&gt;The web UI&lt;/strong&gt; gives you all of this in a browser. Launch the bundled caplet:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;shellscript&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span style=&quot;color: #89B4FA;font-style: italic;&quot;&gt;sudo&lt;/span&gt;&lt;span style=&quot;color: #A6E3A1;&quot;&gt; bettercap -iface eth0 -caplet http-ui&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Then open &lt;code&gt;http://127.0.0.1:80&lt;/code&gt; and log in with the credentials the console
prints. Change the default &lt;code&gt;user/pass&lt;/code&gt; before using it anywhere but localhost.&lt;/p&gt;
&lt;p&gt;The web UI is handy, but it means leaving the terminal and running a browser. If
you live in a terminal, &lt;strong&gt;bettercaptui&lt;/strong&gt; gives you the same live dashboard
without either, pointed at the very same API this caplet exposes.&lt;/p&gt;
&lt;h2 id=&quot;where-to-go-next&quot;&gt;Where to go next&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;BLE:&lt;/strong&gt; &lt;code&gt;ble.recon&lt;/code&gt; scans Bluetooth Low Energy devices, and &lt;code&gt;ble.enum&lt;/code&gt; reads
their characteristics. A good exercise against a smart bulb or fitness band you
own.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;HID:&lt;/strong&gt; the &lt;code&gt;hid&lt;/code&gt; module scans 2.4GHz wireless keyboards and mice and can
inject keystrokes (MouseJacking) with DuckyScript.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Scripting:&lt;/strong&gt; beyond caplets, the JavaScript proxy scripts let you rewrite
HTTP traffic on the fly, and the REST API lets you drive bettercap from your own
tools.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Reading:&lt;/strong&gt; the &lt;a rel=&quot;external&quot; href=&quot;https://www.bettercap.org/modules/&quot;&gt;official module reference&lt;/a&gt;
documents every command, and &lt;code&gt;help &amp;lt;module&amp;gt;&lt;/code&gt; is the same material in the
console.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The through-line: every attack here has a defense, and running them against your
own lab is how you learn both halves. Keep it inside that lab.&lt;/p&gt;
&lt;p&gt;And if the console&#39;s firehose of text wore thin anywhere in this walkthrough,
that&#39;s exactly the itch
&lt;a rel=&quot;external&quot; href=&quot;https://github.com/NPCmillionaire/bettercaptui&quot;&gt;bettercaptui&lt;/a&gt; scratches. Clone
it, point it at your lab, and run the same exercises with live panels instead of
scrollback. Stars, issues, and PRs welcome.&lt;/p&gt;
&lt;hr&gt;
&lt;p&gt;&lt;strong&gt;Source code:&lt;/strong&gt; &lt;a rel=&quot;external&quot; href=&quot;https://github.com/NPCmillionaire/bettercaptui&quot;&gt;bettercaptui on GitHub&lt;/a&gt;&lt;/p&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Burp Suite 101: Reading and Manipulating GET Requests</title>
        <published>2026-08-27T00:00:00+00:00</published>
        <updated>2026-08-27T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/burp-suite-get-requests/"/>
        <id>https://legitbiz.xyz/posts/burp-suite-get-requests/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/burp-suite-get-requests/">&lt;p&gt;Every button click, page load, and &quot;view profile&quot; link on the web
boils down to a plain-text request your browser builds and fires at a
server. Most people never see that text -- the browser assembles it,
sends it, and renders whatever comes back, all invisibly. Burp Suite&#39;s
entire value proposition is removing that invisibility: it sits
between your browser and the internet, shows you the exact request
before it leaves your machine, and lets you change any part of it
before it goes. Once you can see and edit that layer, &quot;the app decided
what to show you&quot; quietly becomes &quot;you decided what to ask for.&quot;&lt;/p&gt;
&lt;h2 id=&quot;what-a-get-request-actually-is&quot;&gt;What a GET request actually is&lt;/h2&gt;
&lt;p&gt;A GET request is one of the HTTP methods browsers use to ask a server
for something, and it&#39;s the one doing the work every time you load a
page, click a link, or follow a bookmark. Stripped down, it looks like
this:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;GET /account/profile?user_id=4471&amp;amp;view=summary HTTP/1.1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;Host: example.com&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;User-Agent: Mozilla/5.0 (...)&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;Cookie: session=a93f...c02e&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;Referer: https://example.com/dashboard&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;Accept: text/html&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Three parts matter most. The &lt;strong&gt;path&lt;/strong&gt; (&lt;code&gt;/account/profile&lt;/code&gt;) tells the
server which resource you want. The &lt;strong&gt;query string&lt;/strong&gt; (everything
after the &lt;code&gt;?&lt;/code&gt;) passes parameters as plain key-value pairs --
&lt;code&gt;user_id=4471&lt;/code&gt; and &lt;code&gt;view=summary&lt;/code&gt; here -- and it&#39;s fully visible in
the URL bar, browser history, and server access logs, which is exactly
why it&#39;s a bad place for anything sensitive. The &lt;strong&gt;headers&lt;/strong&gt; carry
metadata about the request: &lt;code&gt;Cookie&lt;/code&gt; for session state, &lt;code&gt;Referer&lt;/code&gt; for
where you came from, &lt;code&gt;User-Agent&lt;/code&gt; for what client you&#39;re claiming to
be. Unlike POST, a GET request has no body -- everything it sends has
to fit in the path, the query string, or the headers, which is part of
why so much application state ends up sitting in a URL where anyone
looking over your shoulder, or your ISP, or a browser history sync
service, can see it.&lt;/p&gt;
&lt;h2 id=&quot;getting-burp-in-the-middle&quot;&gt;Getting Burp in the middle&lt;/h2&gt;
&lt;p&gt;Burp Suite (the Community edition is free and enough for all of this)
works as an intercepting proxy: your browser sends traffic to Burp
instead of directly to the internet, Burp shows it to you, and only
then forwards it on. Setting that up is three steps:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Point your browser at Burp&#39;s proxy listener.&lt;/strong&gt; By default Burp
listens on &lt;code&gt;127.0.0.1:8080&lt;/code&gt;. Set that as your browser&#39;s HTTP/HTTPS
proxy -- directly in browser settings, or with an extension like
FoxyProxy if you want to flip it on and off without digging through
menus every time.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Install Burp&#39;s CA certificate.&lt;/strong&gt; HTTPS traffic is encrypted
between your browser and the server, so to read it Burp has to
terminate the TLS connection itself and re-encrypt it with its own
certificate -- a deliberate, visible man-in-the-middle that you&#39;re
installing on your own machine, not doing to someone else. Visit
&lt;code&gt;http://burpsuite&lt;/code&gt; while proxied through Burp, download the CA
cert, and add it to your browser or OS trust store. Skip this and
every HTTPS site just throws certificate warnings at you.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Turn on Intercept.&lt;/strong&gt; In the Proxy tab, &quot;Intercept is on&quot; pauses
every request before it leaves your browser, showing you the raw
text and giving you a chance to edit it. Turn it off once you&#39;re
done actively tampering -- left on, it pauses every request on
every site you visit, which gets old fast.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;reading-live-traffic&quot;&gt;Reading live traffic&lt;/h2&gt;
&lt;p&gt;With Intercept off but the proxy still running, browse normally and
watch the &lt;strong&gt;Proxy &amp;gt; HTTP History&lt;/strong&gt; tab fill up with every request your
browser made -- every image, script, API call, and page load, each one
inspectable after the fact. This alone is worth the setup: it&#39;s the
fastest way to see what a page is actually doing under the hood,
including requests to endpoints that never show up in the page source
because JavaScript fired them dynamically. Click any entry and you get
the full raw request on one side, the full raw response on the other.&lt;/p&gt;
&lt;h2 id=&quot;manipulating-the-request&quot;&gt;Manipulating the request&lt;/h2&gt;
&lt;p&gt;This is where Burp stops being a packet sniffer and starts being a
testing tool. Right-click any request in the history and send it to
&lt;strong&gt;Repeater&lt;/strong&gt;, which lets you edit and resend it as many times as you
want, watching the response change with every tweak -- no need to
re-trigger the action in the browser each time.&lt;/p&gt;
&lt;p&gt;A few of the most common things worth trying against a GET request,
assuming you have permission to test the target:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Parameter tampering.&lt;/strong&gt; If a request reads
&lt;code&gt;?user_id=4471&amp;amp;view=summary&lt;/code&gt;, what happens if you change it to
&lt;code&gt;4470&lt;/code&gt;? Or &lt;code&gt;4472&lt;/code&gt;? If the server returns someone else&#39;s account
summary without checking that the logged-in session actually owns
that &lt;code&gt;user_id&lt;/code&gt;, you&#39;ve found an IDOR (Insecure Direct Object
Reference) -- one of the most common and most consequential bugs in
real applications, because it usually takes nothing but changing one
number.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Path and value fuzzing.&lt;/strong&gt; Try unexpected values in numeric fields
(negative numbers, zero, absurdly large numbers, non-numeric
strings), path traversal sequences like &lt;code&gt;../../etc/passwd&lt;/code&gt; in any
parameter that looks like it might reference a filename, and
boolean-ish values (&lt;code&gt;true&lt;/code&gt;/&lt;code&gt;false&lt;/code&gt;, &lt;code&gt;0&lt;/code&gt;/&lt;code&gt;1&lt;/code&gt;) in flags you weren&#39;t
meant to control.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Header manipulation.&lt;/strong&gt; The &lt;code&gt;Referer&lt;/code&gt; and &lt;code&gt;User-Agent&lt;/code&gt; headers are
entirely client-supplied and trivially fake -- if an application
makes an access decision based on either one (&quot;only allow this if
they came from our own site&quot;), that check is worth pressure-testing.
Editing the &lt;code&gt;Cookie&lt;/code&gt; header directly lets you test session handling
without touching browser dev tools at all.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Batch testing with Intruder.&lt;/strong&gt; Once you&#39;ve found one parameter
worth poking, Intruder automates sending the same request with a
list of substituted values -- a range of IDs, a wordlist, a set of
payloads -- and tables the responses so you can spot the one that
behaved differently (a different status code, a different response
length, a different piece of data that shouldn&#39;t be there).&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;None of this requires writing code. It requires noticing that a
request is just text, and that the server has no way of knowing
whether that text came from your browser doing what it was told or
from you doing something else with the same shape.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Client-supplied values are not access control.&lt;/strong&gt; Any check based
on a &lt;code&gt;user_id&lt;/code&gt; in the query string, a &lt;code&gt;Referer&lt;/code&gt; header, or a hidden
form field is a check the client controls entirely -- the fix is
verifying on the server, against the authenticated session, that the
requester is actually allowed to see the resource being asked for,
every time, not just when the request &quot;looks normal.&quot;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Sensitive data doesn&#39;t belong in a GET request at all.&lt;/strong&gt; Query
strings get logged by servers, proxies, and browser history, and
cached by browsers and CDNs, all by design. Anything that
identifies a user or authorizes an action should ride in a POST body
or a header, not a URL parameter, regardless of how convenient the
URL parameter is to implement.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Rate limiting turns manual tampering into an actual barrier
against Intruder.&lt;/strong&gt; A handful of requests per second per session or
IP doesn&#39;t slow down a real user, but it makes automated parameter
sweeps take long enough, and generate enough log noise, to be
noticed and stopped before they finish.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Input validation belongs on the server, as an allowlist, not a
denylist.&lt;/strong&gt; Blocking known-bad patterns (&lt;code&gt;../&lt;/code&gt;, &lt;code&gt;&amp;lt;script&amp;gt;&lt;/code&gt;) only
ever covers the patterns someone thought of in advance. Defining
what a valid &lt;code&gt;user_id&lt;/code&gt; or &lt;code&gt;view&lt;/code&gt; value looks like and rejecting
everything else closes the door on the payloads nobody&#39;s tried yet
too.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;TLS termination in the middle is detectable, and that&#39;s the
point.&lt;/strong&gt; Certificate pinning and HSTS exist specifically so a
Burp-style proxy can&#39;t silently sit between a real user and your
application without the client noticing something&#39;s wrong --
because the same technique that makes testing your own app
possible is exactly what an attacker needs on someone else&#39;s
network.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>You Don&#39;t Need a Hacker. CyberBackgroundChecks Already Doxxed You.</title>
        <published>2026-08-27T00:00:00+00:00</published>
        <updated>2026-08-27T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/data-brokers-already-doxxed-you/"/>
        <id>https://legitbiz.xyz/posts/data-brokers-already-doxxed-you/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/data-brokers-already-doxxed-you/">&lt;p&gt;Doxxing has a reputation as a skill -- something that takes a
motivated stranger, a few hours, and a working knowledge of OSINT
technique to compile your name, address, phone number, relatives, and
property records into one ugly little dossier. That reputation is
mostly wrong, and the reason it&#39;s wrong is a genuinely legal industry
that did the compiling for them years ago and put a price tag on the
printout. Go to cyberbackgroundchecks.com right now, type a common
name plus a state, and you&#39;ll get a results page with age, current and
past addresses, phone numbers, email addresses, relatives and known
associates, and property records, for free, before anyone pays a
cent for the &quot;full report.&quot; Nobody hacked anything. That&#39;s the part
worth sitting with.&lt;/p&gt;
&lt;h2 id=&quot;what-s-actually-sitting-there&quot;&gt;What&#39;s actually sitting there&lt;/h2&gt;
&lt;p&gt;CyberBackgroundChecks -- and the several hundred sites that work
exactly like it -- pulls from court records, property/real estate
filings, voter registration data, and a layer of purchased data from
other data brokers and social media scraping, then packages the
result as a &quot;background check.&quot; It isn&#39;t hiding this; it&#39;s in their
own privacy policy. What makes it feel like doxxing rather than
&quot;public records&quot; is presentation: instead of a dozen scattered county
databases you&#39;d have to know existed and query one at a time, it&#39;s
one search box, one page, your whole footprint assembled for you by
someone who never had to explain why they wanted it.&lt;/p&gt;
&lt;p&gt;The legal fine print that makes this all permissible is the same
sentence you&#39;ll find at the bottom of every site like it: not to be
used for FCRA-covered purposes like employment, tenant, or credit
decisions. That disclaimer exists because the Fair Credit Reporting
Act imposes real accuracy and dispute obligations on anything used to
decide whether someone gets a job or an apartment -- and these sites
have opted out of that entire regulatory category by simply telling
you not to use it that way, while doing nothing to stop you.&lt;/p&gt;
&lt;h2 id=&quot;how-your-data-got-there-in-the-first-place&quot;&gt;How your data got there in the first place&lt;/h2&gt;
&lt;p&gt;Nobody hands their address to CyberBackgroundChecks directly. It
arrives sideways, through dozens of small, individually reasonable
disclosures that were never supposed to add up to a public profile:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Registering to vote, which in most states creates a public record
including your address by default.&lt;/li&gt;
&lt;li&gt;Buying property, which generates a county deed record that&#39;s
public unless you specifically title it through a trust or LLC.&lt;/li&gt;
&lt;li&gt;Any court filing you&#39;re party to -- divorce, small claims, an
eviction, a lawsuit either direction -- all of which are public
record in the overwhelming majority of jurisdictions.&lt;/li&gt;
&lt;li&gt;The loyalty program, warranty card, or &quot;10% off your first order&quot;
email signup that quietly sold your info downstream to a data
broker you&#39;ve never heard of, who sold it to another one, who sold
it to the one that finally built the page with your name on it.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Data brokers aggregate all of it, cross-reference it against itself to
resolve &quot;is this the same person,&quot; and resell the merged profile.
CyberBackgroundChecks doesn&#39;t need to do original research. It&#39;s a
retail storefront in front of a wholesale data supply chain that&#39;s
been running for years.&lt;/p&gt;
&lt;h2 id=&quot;getting-yourself-off-of-it&quot;&gt;Getting yourself off of it&lt;/h2&gt;
&lt;p&gt;CyberBackgroundChecks has an actual opt-out process, and it&#39;s not
buried as deep as you&#39;d expect: their removal page is at
&lt;code&gt;cyberbackgroundchecks.com/removal&lt;/code&gt; (also linked from the &quot;Privacy&quot;
link in the site footer, under the &quot;Your Rights&quot; section of the
privacy policy). The process is: submit your name and email, confirm
via a verification email within 24 hours, then complete a &quot;record
suppression&quot; form matching your listing. Removal is generally
processed within about three days once verification is complete.&lt;/p&gt;
&lt;p&gt;As of 2026, California residents have a meaningfully better option:
DROP, the state&#39;s Delete Request and Opt-Out Platform, went live
January 1, 2026, and lets you file a single deletion request that
every data broker registered in California -- over 500 of them -- is
legally required to honor, instead of filing this same removal form
site by site by site. Enforcement has teeth behind it: starting August
1, 2026, registered brokers must check DROP for pending requests at
least every 45 days and process any match, or face fines. It&#39;s the
first system of its kind in the country. It&#39;s also, right now,
California-only.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Removal isn&#39;t a fix, it&#39;s a cycle.&lt;/strong&gt; These sites rebuild profiles
the moment new source data surfaces -- a new address, a new court
filing, a new data-broker feed -- with no memory that you opted out
before. Practically every removal guide for sites like this ends
with &quot;check back periodically,&quot; because that&#39;s the actual expected
outcome, not an edge case.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;DROP treats the symptom, not the source.&lt;/strong&gt; Even a perfect DROP
deletion doesn&#39;t stop your name from re-entering a voter file,
property record, or court docket next month -- and once it does,
the next broker crawl picks it back up and the 45-day clock starts
over. It&#39;s a recurring deletion right, not a standing shield.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;It&#39;s a state-by-state patchwork.&lt;/strong&gt; DROP only binds brokers with
respect to California residents&#39; data; if you live anywhere else,
you&#39;re still filing removal forms one site at a time, or paying a
service like DeleteMe or Incogni to do it for you on a subscription,
which is itself an admission that the manual process doesn&#39;t scale.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The real leverage is upstream.&lt;/strong&gt; Property held through a trust or
LLC instead of your personal name, address confidentiality programs
that some states offer to domestic violence survivors and others at
elevated risk, and skipping the &quot;enter your info for a discount&quot;
forms all reduce what enters the pipeline in the first place --
which does more long-term work than deleting any single downstream
listing, because the listing was never the source of the problem.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Building a DIY WiFi Pineapple Out of a Raspberry Pi</title>
        <published>2026-08-27T00:00:00+00:00</published>
        <updated>2026-08-27T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/diy-wifi-pineapple-raspberry-pi/"/>
        <id>https://legitbiz.xyz/posts/diy-wifi-pineapple-raspberry-pi/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/diy-wifi-pineapple-raspberry-pi/">&lt;p&gt;The WiFi Pineapple is Hak5&#39;s commercial rogue-AP platform -- a small
box with two radios and a web dashboard for running evil-twin APs,
Karma/MANA attacks, deauth, and captive portals. Almost all of that
is just Linux plumbing: &lt;code&gt;hostapd&lt;/code&gt;, &lt;code&gt;dnsmasq&lt;/code&gt;, &lt;code&gt;iptables&lt;/code&gt;, and the
aircrack-ng suite, wired together. A Raspberry Pi with the right USB
WiFi adapter reproduces the core functionality for the cost of the
hardware, and building it by hand instead of flashing someone else&#39;s
firmware means you actually understand what each piece is doing.&lt;/p&gt;
&lt;p&gt;Everything below is for testing your own network, an isolated lab, or
an engagement with signed, scoped authorization. A rogue AP that
intercepts or manipulates traffic from devices that aren&#39;t yours --
or people who haven&#39;t agreed to it -- crosses into wiretapping and
CFAA territory immediately. This is the same category as lockpicks:
legal to own and practice with, illegal to point at a lock that isn&#39;t
yours.&lt;/p&gt;
&lt;h2 id=&quot;what-you-need&quot;&gt;What you need&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Raspberry Pi 4B or 5&lt;/strong&gt; (2GB+). Running &lt;code&gt;hostapd&lt;/code&gt;, &lt;code&gt;dnsmasq&lt;/code&gt;, and
packet capture at once wants the extra headroom over a Pi Zero.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A USB WiFi adapter with monitor mode and packet injection
support.&lt;/strong&gt; The Pi&#39;s onboard radio doesn&#39;t do injection -- this is
non-negotiable. Atheros AR9271 and Realtek RTL8812AU chipsets
(Alfa AWUS036NHA, AWUS036ACH) are the known-good picks, same as in
the &lt;a href=&quot;/posts/aircrack-ng-wpa2-handshake&quot;&gt;aircrack-ng post&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;A second radio -- either the Pi&#39;s built-in WiFi or a second USB
adapter -- so one interface can serve the rogue AP while the other
stays free for scanning or injection.&lt;/li&gt;
&lt;li&gt;microSD card (32GB+), a powered USB hub if you&#39;re running a
high-gain adapter (they brown out the Pi&#39;s ports), and a battery
pack for field use.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;flashing-and-base-setup&quot;&gt;Flashing and base setup&lt;/h2&gt;
&lt;p&gt;Raspberry Pi OS Lite (64-bit) or Kali ARM both work; Kali ships most
of the wireless tooling preinstalled. Enable SSH before first boot so
you can work headless, then:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo apt update &amp;amp;&amp;amp; sudo apt full-upgrade -y&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo apt install -y hostapd dnsmasq aircrack-ng iptables-persistent \&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;                     dhcpcd5 bridge-utils tcpdump macchanger&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Confirm the USB adapter actually supports monitor mode:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo airmon-ng&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;iw list | grep -A 10 &amp;quot;Supported interface modes&amp;quot;&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;h2 id=&quot;standing-up-the-rogue-ap&quot;&gt;Standing up the rogue AP&lt;/h2&gt;
&lt;p&gt;Identify interfaces with &lt;code&gt;iwconfig&lt;/code&gt; -- typically &lt;code&gt;wlan0&lt;/code&gt; (onboard)
and &lt;code&gt;wlan1&lt;/code&gt; (USB adapter) -- then stop the services so you can
hand-configure them:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo systemctl stop hostapd dnsmasq&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo rfkill unblock wifi&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;code&gt;/etc/hostapd/hostapd.conf&lt;/code&gt;:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;interface=wlan1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;driver=nl80211&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;ssid=Free_Public_WiFi&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;hw_mode=g&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;channel=6&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;wmm_enabled=0&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;macaddr_acl=0&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;auth_algs=1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;ignore_broadcast_ssid=0&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo sed -i &amp;#39;s/#DAEMON_CONF=.*/DAEMON_CONF=&amp;quot;\/etc\/hostapd\/hostapd.conf&amp;quot;/&amp;#39; /etc/default/hostapd&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;code&gt;/etc/dnsmasq.conf&lt;/code&gt;:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;interface=wlan1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;dhcp-range=10.0.0.10,10.0.0.250,255.255.255.0,12h&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;dhcp-option=3,10.0.0.1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;dhcp-option=6,10.0.0.1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;server=8.8.8.8&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;log-queries&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;log-dhcp&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo ip addr add 10.0.0.1/24 dev wlan1&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;h2 id=&quot;routing-it-to-the-internet&quot;&gt;Routing it to the internet&lt;/h2&gt;
&lt;p&gt;If you want connected clients to get real internet (useful for
credential-harvesting or MITM audits where a dead-end AP would tip
off the target), NAT &lt;code&gt;wlan1&lt;/code&gt; out through &lt;code&gt;wlan0&lt;/code&gt;:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo sysctl -w net.ipv4.ip_forward=1&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo iptables -t nat -A POSTROUTING -o wlan0 -j MASQUERADE&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo iptables -A FORWARD -i wlan0 -o wlan1 -m state --state RELATED,ESTABLISHED -j ACCEPT&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo iptables -A FORWARD -i wlan1 -o wlan0 -j ACCEPT&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo netfilter-persistent save&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Then bring the AP up:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo systemctl unmask hostapd&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;sudo systemctl start hostapd dnsmasq&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;At this point &lt;code&gt;wlan1&lt;/code&gt; is an open AP handing out DHCP leases and
routing clients through &lt;code&gt;wlan0&lt;/code&gt; -- the evil-twin core of what a
Pineapple does out of the box.&lt;/p&gt;
&lt;h2 id=&quot;layering-on-the-actual-pineapple-style-attacks&quot;&gt;Layering on the actual Pineapple-style attacks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Karma/MANA&lt;/strong&gt; (auto-responding to client probe requests so nearby
devices connect without a prompt) -- the &lt;a rel=&quot;external&quot; href=&quot;https://github.com/sensepost/mana&quot;&gt;mana
toolkit&lt;/a&gt; is the standard
implementation.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Handshake capture&lt;/strong&gt; for auditing your own network&#39;s passphrase
strength -- &lt;code&gt;airodump-ng&lt;/code&gt; and &lt;code&gt;aireplay-ng&lt;/code&gt;, covered in detail in
the &lt;a href=&quot;/posts/aircrack-ng-wpa2-handshake&quot;&gt;aircrack-ng post&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Captive-portal credential audits&lt;/strong&gt; -- &lt;code&gt;wifiphisher&lt;/code&gt; automates a
fake login page; legitimate use here is limited to authorized
phishing-awareness testing with a signed scope.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A web dashboard instead of raw config files&lt;/strong&gt; --
&lt;a rel=&quot;external&quot; href=&quot;https://github.com/xtr4nge/FruityWiFi&quot;&gt;FruityWiFi&lt;/a&gt; is an
open-source project explicitly modeled on the Pineapple&#39;s plugin
architecture, and drops on top of everything above.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;where-this-goes-wrong-in-practice&quot;&gt;Where this goes wrong in practice&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;The onboard radio can&#39;t inject.&lt;/strong&gt; The single most common failure
mode for people following Pineapple tutorials on a Pi is trying to
do all of this with the built-in WiFi chip. It&#39;ll join networks
fine and it&#39;ll even host an AP, but it silently can&#39;t do
monitor-mode packet injection -- you need the external adapter for
anything beyond serving a plain rogue AP.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Signal bleed outside the intended range.&lt;/strong&gt; An open &quot;Free_Public_
WiFi&quot; SSID doesn&#39;t respect your test boundary -- it advertises to
every device in radio range, not just the ones you meant to test.
Testing in a Faraday bag or a shielded room is the only way to keep
an authorized test from touching networks and devices it wasn&#39;t
scoped for.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;DHCP/DNS collisions on a live network.&lt;/strong&gt; Running &lt;code&gt;dnsmasq&lt;/code&gt; on a
network segment that already has a DHCP server produces a race
between the two -- clients get inconsistent leases and legitimate
users notice connectivity breaking. This is a real risk during an
engagement and a good way to turn an authorized test into an
unplanned outage.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Karma-style auto-connect is exactly what modern OSes try to
block.&lt;/strong&gt; iOS, Android, and recent Windows/macOS builds increasingly
randomize MAC addresses and refuse to auto-join open networks by
SSID alone, which is why Karma-era attacks work less reliably than
they did five years ago -- a captive portal with a plausible pretext
does more work now than the auto-connect trick by itself.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No authorization, no defense.&lt;/strong&gt; None of the mitigations above
change the legal exposure of running this against a network or
people who haven&#39;t consented. Written scope, a defined time window,
and logging (&lt;code&gt;tcpdump -w capture.pcap&lt;/code&gt; for the whole session) are
what separates a pentest from a crime with the exact same packet
capture.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Inside the Antifraud AI: FICO Falcon, Visa VAA, and Mastercard&#39;s Decision Intelligence</title>
        <published>2026-08-27T00:00:00+00:00</published>
        <updated>2026-08-27T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/inside-the-antifraud-ai/"/>
        <id>https://legitbiz.xyz/posts/inside-the-antifraud-ai/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/inside-the-antifraud-ai/">&lt;p&gt;Every card swipe, tap, or checkout gets scored by a fraud model before
it&#39;s approved -- usually in under 300 milliseconds. Three systems
handle the overwhelming majority of that scoring worldwide: &lt;strong&gt;FICO
Falcon Fraud Manager&lt;/strong&gt;, licensed by most issuing banks; &lt;strong&gt;Visa
Advanced Authorization&lt;/strong&gt;, run at the network level on every Visa
transaction; and &lt;strong&gt;Mastercard Decision Intelligence&lt;/strong&gt; (built on the
acquired Brighterion stack). [[the-fraud-score]] covered the score
itself and what feeds it from a merchant/processor angle -- this is
about the specific models sitting upstream of that, at the issuer and
network level, and how they&#39;re actually built.&lt;/p&gt;
&lt;h2 id=&quot;the-three-layer-stack&quot;&gt;The three-layer stack&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Rules engines&lt;/strong&gt; -- the oldest, still-running layer. Hard blocks:
certain countries, dollar ceilings, blacklisted merchant category
codes. Fast and explainable, but static -- anyone who sees a decline
reason enough times can map the rule.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Supervised models&lt;/strong&gt; -- gradient-boosted trees and neural nets
trained on billions of transactions labeled fraud/not-fraud after
the fact (via chargebacks and confirmed fraud reports). Output is a
continuous score, not a yes/no, which is what lets an issuer route
a mid-range score to a step-up challenge (OTP, 3DS) instead of an
outright decline.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Unsupervised/graph models&lt;/strong&gt; -- anomaly detection and network
analysis that don&#39;t need a labeled fraud example to work. This is
the layer that catches new patterns before enough chargebacks exist
to train a supervised model on them, and it&#39;s the layer that links
a fraud ring together by shared device IDs, IPs, or shipping
addresses across accounts that look unrelated transaction-by-
transaction.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Mastercard&#39;s Decision Intelligence leans hardest on the graph layer --
its whole pitch is modeling relationships between accounts, not just
scoring accounts individually. Visa&#39;s VAA runs network-wide, so it
sees patterns no single issuing bank could: a BIN range or device
fingerprint showing up in confirmed fraud at one bank gets weighted
into risk scores at every other bank on the network within the same
scoring cycle.&lt;/p&gt;
&lt;h2 id=&quot;what-actually-feeds-the-model&quot;&gt;What actually feeds the model&lt;/h2&gt;
&lt;p&gt;Same categories regardless of vendor:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Identity/device age and stability&lt;/strong&gt; -- how long has this card,
email, phone, or device been seen, and how consistently.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Velocity&lt;/strong&gt; -- transaction count and dollar volume per card,
device, IP, or shipping address in a rolling window. The strongest
single signal, because compartmentalizing every session across
fake identities is expensive to do perfectly.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Network reputation&lt;/strong&gt; -- has this card BIN, device fingerprint, or
IP shown up in confirmed fraud anywhere else in the cross-issuer
dataset. This is the entire value proposition of a network-level
system like VAA over a single bank&#39;s in-house model: one bank sees
its own fraud, the network sees everyone&#39;s.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Behavioral biometrics&lt;/strong&gt; -- typing cadence, swipe pressure, how a
phone is held during a mobile-app session. Newer signal, harder to
fake at scale than a device ID.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Every signal is individually noisy -- a VPN, a shared family card, a
person traveling all trip flags that mean nothing on their own. The
model&#39;s entire job is combining weak signals into something with
better precision than any single one.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;p&gt;None of this is secret from the people trying to beat it, and the
patterns below are exactly why issuers treat model thresholds as
closely guarded and rotate them on a schedule rather than a fixed
config:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Velocity smoothing&lt;/strong&gt; -- instead of many fast small charges (an
obvious spike), spreading test transactions across time and
merchants keeps every individual window under the threshold.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Identity aging&lt;/strong&gt; -- synthetic or warmed-up identities built weeks
or months ahead of use have no anomaly to detect against, because
the model is looking for deviation from a baseline that was
deliberately built to look normal.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Device/location normalization&lt;/strong&gt; -- matching a transaction&#39;s
apparent IP and device fingerprint to the cardholder&#39;s usual
geography defeats the network-reputation signal specifically,
since that signal depends on the fingerprint looking foreign or
previously flagged.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Targeting the human layer instead of the model&lt;/strong&gt; -- SIM-swaps,
phishing, and call-center pretexting produce a session that&#39;s
cryptographically and behaviorally identical to the real
cardholder, which is precisely the case none of the transaction-
level signals above are built to catch. This is why account
takeover, not card-present fraud, is the fastest-growing category
issuers report.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Threshold probing&lt;/strong&gt; -- running small transactions specifically to
map where a given issuer&#39;s decline line sits, then staying just
under it. Organized rings do this systematically; it&#39;s the reason
issuers now rotate thresholds and blend in randomized manual review
even for scores that would otherwise auto-clear.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The practical upshot: none of these systems are a wall, they&#39;re a
tuned filter balancing fraud losses against false declines, and any
filter with a threshold can eventually be mapped by someone patient
enough to probe it. The industry&#39;s current bet on staying ahead is
leaning harder on the signals that are actually expensive to fake --
graph relationships across accounts, and behavioral biometrics -- over
any single transaction-level data point.&lt;/p&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Snitches Get Statistics: Why (Almost) Nobody Goes to Trial</title>
        <published>2026-08-27T00:00:00+00:00</published>
        <updated>2026-08-27T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/snitches-get-statistics/"/>
        <id>https://legitbiz.xyz/posts/snitches-get-statistics/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/snitches-get-statistics/">&lt;p&gt;Every fraud case you&#39;ve ever read about ends the same way in the last
paragraph of the press release: &quot;pleaded guilty.&quot; Not &quot;was convicted
at trial by a jury of their peers,&quot; which is the version everyone&#39;s
mental model of the justice system is still running on. Pleaded
guilty. Almost always. Almost immediately. And in a multi-defendant
fraud ring -- which is most of them, because fraud this size takes
more than one person -- almost always in a specific, predictable
order, driven by a dynamic that never makes it into the press release
at all: somebody flipped, and once they did, the math for everyone
else still holding out got a lot worse.&lt;/p&gt;
&lt;h2 id=&quot;the-trial-is-basically-theater-at-this-point&quot;&gt;The trial is basically theater at this point&lt;/h2&gt;
&lt;p&gt;Start with the baseline, because it&#39;s more extreme than most people
guess. In fiscal year 2024, 97% of sentenced federal defendants got
there by pleading guilty -- meaning roughly 3% of cases went to trial
at all. That&#39;s not a fraud-specific number, that&#39;s everything, and
it&#39;s been trending that direction for decades: thirty years ago,
about 20% of people who got arrested chose trial. Today it&#39;s under
3%, and has been for most of the last decade. Fraud and financial
crime cases track this closely or run a little higher, because
they&#39;re document-heavy, paper-trail-heavy, and about the least
sympathetic thing you can put in front of a jury next to &quot;I panicked.&quot;&lt;/p&gt;
&lt;p&gt;None of this means everyone pleading guilty actually did it, and
that&#39;s the part that should bother you more than it probably does.
It means the system built a machine that is extremely good at
extracting a guilty plea and only incidentally good at extracting the
truth. Those turned out not to be the same design goal.&lt;/p&gt;
&lt;h2 id=&quot;the-number-that-actually-explains-it&quot;&gt;The number that actually explains it&lt;/h2&gt;
&lt;p&gt;The mechanism doing most of the work here isn&#39;t complicated, it&#39;s
just rarely said out loud: pleading guilty is &lt;em&gt;cheaper&lt;/em&gt;, and the
discount is enormous. The National Association of Criminal Defense
Lawyers spent years documenting this in a 2018 report bluntly titled
&quot;The Trial Penalty,&quot; and the numbers are not subtle -- federal
sentences after a trial conviction run, on average, roughly three
times longer than the sentence the same person could have gotten by
pleading, and in some case types the gap runs eight to ten times.
Averaged across federal felonies generally, that&#39;s something like a
seven-year difference in sentence length for exercising a
constitutional right instead of waiving it.&lt;/p&gt;
&lt;p&gt;Nobody calls this a penalty for going to trial, officially. Officially
it&#39;s &quot;acceptance of responsibility&quot; credit you forfeit, and mandatory
minimums you&#39;re no longer shielded from once a jury -- rather than a
plea agreement -- decides your fate. Functionally, it&#39;s a tax on
making the government prove its case, and it&#39;s steep enough that
plenty of innocent people do the math and plead anyway, because losing
at trial doesn&#39;t cost you a little more than pleading guilty, it costs
you years more. This is the pressure sitting on every defendant before
a single codefendant has said a word to a prosecutor.&lt;/p&gt;
&lt;h2 id=&quot;where-the-codefendant-comes-in&quot;&gt;Where the codefendant comes in&lt;/h2&gt;
&lt;p&gt;Add a second, third, or fifteenth defendant to the same indictment,
and you&#39;ve built a prisoner&#39;s dilemma with a first-mover bonus
attached. The prosecutor doesn&#39;t need all of them to cooperate. They
need one -- ideally the one closest to the middle of the conspiracy,
with documents and a working memory of who did what. Once that person
signs a cooperation agreement, the value of a second cooperator drops,
and the value of a third drops close to zero, because the government
already has an insider witness and doesn&#39;t need to trade a sentence
reduction for information it can already prove. Defense attorneys who
work multi-defendant fraud cases describe it, without much
disagreement, as a literal race: timing matters more than how guilty
you actually are relative to your codefendants, because the reward is
almost entirely a function of arrival order.&lt;/p&gt;
&lt;p&gt;The going rate for winning that race is visible in U.S. Sentencing
Commission data on &quot;substantial assistance&quot; departures under
guideline section 5K1.1 -- the formal mechanism for rewarding
cooperation with a lighter sentence. Nationally, fraud offenders make
up roughly 13.5% of everyone who gets one, and the departure itself
isn&#39;t symbolic: it knocks an average of 52.6% off the bottom of the
defendant&#39;s sentencing guideline range. That&#39;s the size of the prize
for being the one who talks. It&#39;s also not guaranteed money in the
bank -- cooperation agreements get revoked for breach in something
like 11% of cases, which is prosecutor-speak for &quot;we decided you
weren&#39;t actually being straight with us,&quot; and at that point you&#39;ve
already confessed on the record with nothing to show for it.&lt;/p&gt;
&lt;h2 id=&quot;what-this-looks-like-in-an-actual-ring&quot;&gt;What this looks like in an actual ring&lt;/h2&gt;
&lt;p&gt;The COVID-era PPP loan fraud cases are a good place to watch this play
out in public, because there were so many of them and DOJ press
releases are unusually specific about the sequence. In one Houston
ring, Amir Aqeel was sentenced to 15 years for leading a scheme that
helped at least 14 other people file more than 75 fraudulent PPP loan
applications worth roughly $20 million. All six charged codefendants
in that case pleaded guilty -- nobody went to trial, and the sentences
scaled down sharply the further you were from the center of the
conspiracy, from Aqeel&#39;s 15 years down to 18-month terms for people
who submitted fewer applications and cooperated earlier. In a separate
Atlanta-based PPP ring, by the time the twelfth and final defendant
was charged, all eleven people charged before him -- nine of them
business owners -- had already pleaded guilty or been convicted. At
that point, defendant twelve isn&#39;t deciding whether to fight a case.
He&#39;s deciding how much cooperation credit is still on the table after
eleven other people already told the government everything it needed
to know.&lt;/p&gt;
&lt;h2 id=&quot;the-honest-answer-nobody-actually-publishes-this-number&quot;&gt;The honest answer: nobody actually publishes this number&lt;/h2&gt;
&lt;p&gt;Here&#39;s the part I&#39;d be lying if I dressed up as more solid than it
is: there is no government agency, and no dataset I could find, that
tracks &quot;percentage of guilty pleas caused by a codefendant flipping&quot;
as its own tidy statistic. It&#39;s not that anyone&#39;s hiding it -- it&#39;s
that a guilty plea doesn&#39;t record its own motive. The plea agreement
says the defendant did it and is accepting responsibility; it doesn&#39;t
say &quot;and also my co-conspirator cut a deal eight weeks ago and gutted
my leverage.&quot; Prosecutors have zero incentive to spell that out in a
press release when &quot;pleaded guilty following an investigation&quot; does
the job just as well from their side.&lt;/p&gt;
&lt;p&gt;What you get instead, and what this whole post has been assembling,
is a stack of adjacent numbers that all point the same direction: a
97% overall plea rate, a trial penalty steep enough to make pleading
rational even for people who might win, a cooperation-reward system
that pays out real time off a sentence and is disproportionately used
in fraud cases, and real multi-defendant fraud rings where the guilty
pleas visibly cascade in order once the first person talks. Stack
those up and the missing number -- how much of the 97% is specifically
downstream of a codefendant&#39;s decision to cooperate -- is clearly
large. It&#39;s just not a number anyone was ever going to publish,
because nobody involved benefits from saying it plainly.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;The penalty doesn&#39;t discriminate by guilt.&lt;/strong&gt; A trial penalty that
can add years to a sentence pressures innocent and guilty defendants
identically -- it&#39;s a function of what you&#39;re risking by making the
government work for a conviction, not a function of what you
actually did. Multiple exoneration studies have flagged false guilty
pleas as a real, non-trivial category specifically because of this
math, not as an edge case.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&quot;Substantial assistance&quot; is a bet, not a receipt.&lt;/strong&gt; Roughly 1 in 9
cooperation agreements gets revoked for an alleged breach, and a
defendant who&#39;s already confessed on the record has very little
leverage left to argue about it once that happens.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Timing beats accuracy.&lt;/strong&gt; The reward structure pays out based on
who reaches the prosecutor&#39;s office first, not who was most
culpable or whose information is most reliable -- which means the
first cooperator&#39;s version of events often becomes the government&#39;s
working theory of the whole case before anyone can meaningfully
challenge it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;This dynamic is federal-strength by design.&lt;/strong&gt; Mandatory minimums
and sentencing guidelines are what make the trial penalty large
enough to force the decision; state-level fraud prosecutions, with
generally lighter and more judge-discretionary sentencing, produce
the same race-to-cooperate incentive but with a smaller prize
attached, which is part of why federal prosecutors are the ones who
get first pick of multi-defendant financial crime cases whenever
jurisdiction is contested.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>The Only Article About OPSEC You Need to Know in the AI Age</title>
        <published>2026-08-27T00:00:00+00:00</published>
        <updated>2026-08-27T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/the-only-article-about-opsec-you-need-to-know-in-the-ai-age/"/>
        <id>https://legitbiz.xyz/posts/the-only-article-about-opsec-you-need-to-know-in-the-ai-age/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/the-only-article-about-opsec-you-need-to-know-in-the-ai-age/">&lt;p&gt;Most personal-security advice is still written for 2015: use a strong
password, don&#39;t click links, maybe get a VPN. None of that is wrong,
exactly, but it misses what actually changed. The realistic threats to
an ordinary person in 2026 are a stalker or ex with $30 and a data
broker site, a SIM-swap crew targeting your phone number specifically
because it&#39;s tied to your bank, and AI-assisted scams that can clone a
voice from a ten-second clip or generate a video call convincing enough
to fool a family member. Good OPSEC now means understanding what each
tool in your pocket actually protects against -- and what it doesn&#39;t.&lt;/p&gt;
&lt;h2 id=&quot;start-with-the-phone-because-it-s-the-whole-attack-surface&quot;&gt;Start with the phone, because it&#39;s the whole attack surface&lt;/h2&gt;
&lt;p&gt;Your phone number is the master key to more of your life than your
password is. It&#39;s the recovery method for your email, which is the
recovery method for everything else. The single highest-leverage move
most people can make is treating the phone number itself as a
liability:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Enable a carrier PIN/passcode for SIM changes&lt;/strong&gt;, not just account
login. AT&amp;amp;T, T-Mobile, and Verizon all offer a port-out or SIM-swap
PIN that&#39;s separate from your account password -- set it, because a
SIM swap doesn&#39;t need your password, it needs a call center agent
convinced they&#39;re talking to you.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Move critical 2FA off SMS entirely.&lt;/strong&gt; An authenticator app (or a
hardware key like a YubiKey for accounts that support it) can&#39;t be
intercepted by a SIM swap, because it doesn&#39;t depend on your carrier
at all.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Use Apple&#39;s Advanced Data Protection and Lockdown Mode&lt;/strong&gt; if you&#39;re
on iOS. Advanced Data Protection extends end-to-end encryption to
iCloud backups, photos, and notes -- without it, Apple holds a key
to most of your iCloud data and can hand it over under legal
process. Lockdown Mode is a much blunter tool (it breaks link
previews, some attachments, and other conveniences) built
specifically for people who are realistic, plausible targets of
sophisticated spyware -- journalists, activists, executives -- not
something most people need running day to day.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;facetime-imessage-and-what-encrypted-actually-covers&quot;&gt;FaceTime, iMessage, and what &quot;encrypted&quot; actually covers&lt;/h2&gt;
&lt;p&gt;FaceTime and iMessage are end-to-end encrypted in transit -- Apple
can&#39;t read the content of a call or a blue-bubble message, and neither
can anyone intercepting the network traffic. That&#39;s real, and it&#39;s
more than most people&#39;s default assumption gives it credit for. But
end-to-end encryption has a boundary, and knowing where it sits
matters more than the headline:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;It protects the pipe, not the endpoints.&lt;/strong&gt; If the recipient&#39;s
device is compromised, backed up unencrypted, or the recipient just
screenshots the conversation, encryption in transit did nothing.
This is also exactly how AI-generated impersonation scams work now:
the call itself may be genuinely end-to-end encrypted and still be
someone using a cloned voice or a deepfaked video feed on the other
end. Encryption confirms nobody tampered with the signal between you
two -- it says nothing about who &quot;you two&quot; actually are.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Green bubbles (SMS/RCS to Android) are not end-to-end encrypted&lt;/strong&gt;
the same way -- RCS added encryption for Android-to-Android in 2024,
but cross-platform iPhone-to-Android messages still fall back to
unencrypted SMS/MMS. If the content matters, use a dedicated
end-to-end app (iMessage, Signal, WhatsApp) on both ends rather than
assuming any blue/green bubble is equally protected.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;iCloud backup is the actual gap.&lt;/strong&gt; Without Advanced Data
Protection turned on, a full iCloud backup of your messages and call
history sits on Apple&#39;s servers in a form Apple itself can decrypt
-- meaning the encryption you rely on during the call doesn&#39;t
necessarily follow the transcript into storage.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;vpns-what-they-hide-what-they-don-t&quot;&gt;VPNs: what they hide, what they don&#39;t&lt;/h2&gt;
&lt;p&gt;A VPN is the most oversold consumer security product that still
happens to be worth using. It does one thing well and gets credit for
ten things it doesn&#39;t do:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;What it actually does:&lt;/strong&gt; encrypts your traffic between your
device and the VPN provider&#39;s server, and hides your IP address from
the sites you visit (they see the VPN&#39;s IP, not yours). This is
genuinely useful on untrusted networks -- airport wifi, a coffee
shop -- where the previous hop would otherwise see your unencrypted
traffic.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;What it doesn&#39;t do:&lt;/strong&gt; make you anonymous. The VPN provider itself
can see everything your ISP used to see, which is why the provider&#39;s
no-logs policy is the entire product -- not the encryption, which
every VPN has. A provider that logs and later gets subpoenaed or
breached has just handed over the same data your ISP would have. Pick
based on independently audited no-logs claims and jurisdiction, not
marketing copy.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;What it never touches:&lt;/strong&gt; anything after the traffic decrypts.
Logging into Google or Facebook over a VPN still tells Google and
Facebook it&#39;s you, because you authenticated with your own account.
A VPN hides your network path, not your identity once you sign in to
something that already knows who you are.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;the-ai-age-threats-the-old-advice-doesn-t-cover&quot;&gt;The AI-age threats the old advice doesn&#39;t cover&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Voice cloning from public audio.&lt;/strong&gt; A few seconds of audio from a
voicemail greeting, a social media video, or a public talk is enough
for current voice-cloning tools to produce a convincing clone. The
practical defense isn&#39;t detecting the clone -- most people can&#39;t,
reliably -- it&#39;s having an out-of-band verification habit with
family (a agreed code word, or simply hanging up and calling back on
a known number) for anything involving money or urgency, the same
way banks trained people to distrust &quot;your card has been frozen,
call this number&quot; calls.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Real-time deepfake video on calls.&lt;/strong&gt; Live face-swap and voice
conversion tools have gotten cheap and fast enough to run during an
actual video call, not just a pre-recorded clip. The tell is
usually still there if you look for it -- lighting that doesn&#39;t
quite track head movement, audio slightly desynced from lip
movement, a refusal to turn the head to profile -- but the honest
answer is this arms race favors the attacker right now, which is
why verification should depend on something other than &quot;I saw their
face,&quot; like a shared secret or callback.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;AI-written phishing and pretexting.&lt;/strong&gt; The old advice (&quot;look for
bad grammar and spelling&quot;) is dead; a model writes fluent, well-
targeted text by default now. What hasn&#39;t changed is that urgency
plus a request to act outside your normal channel (a new payment
method, a new phone number, &quot;don&#39;t tell anyone yet&quot;) is still the
actual signal, regardless of how polished the message is.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Perfect tool hygiene doesn&#39;t cover social engineering.&lt;/strong&gt; A SIM PIN,
Advanced Data Protection, and an audited VPN do nothing against a
scam that convinces you, correctly authenticated, to send money or
hand over an OTP code yourself. Most successful compromises in 2026
don&#39;t break encryption -- they route around it through a person.
Verification habits matter more than any single product.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Metadata survives every one of these tools.&lt;/strong&gt; A VPN hides IP, not
the fact that you and someone else both used Signal at 2am for
twenty minutes. iMessage hides content, not who you talked to and
for how long. If who you&#39;re talking to is itself sensitive, content
encryption alone doesn&#39;t cover it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The weakest link is whichever account still uses SMS 2FA or a
reused password&lt;/strong&gt;, not the one you hardened. Attackers don&#39;t beat
your best defense, they find the one you forgot to apply it to --
an old email account, a &quot;just in case&quot; login you haven&#39;t touched in
years, a smart-home app tied to the same phone number as your bank.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;None of this is static.&lt;/strong&gt; RCS encryption, Advanced Data
Protection, and carrier SIM-lock policy have all meaningfully
changed in the last two years alone. Anything written about &quot;the&quot;
right settings is a snapshot, not a permanent answer -- check the
current state before assuming last year&#39;s guide still applies.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Capturing and Cracking a WPA2 Handshake with Aircrack-ng</title>
        <published>2026-08-26T00:00:00+00:00</published>
        <updated>2026-08-26T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/aircrack-ng-wpa2-handshake/"/>
        <id>https://legitbiz.xyz/posts/aircrack-ng-wpa2-handshake/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/aircrack-ng-wpa2-handshake/">&lt;p&gt;You just stood up a new AP -- a spare router for a guest network, a
lab bench box, whatever -- and picked a passphrase for it. The only
honest way to find out whether that passphrase is any good is to
attack it the same way anyone else would. That&#39;s the whole job of
&lt;code&gt;aircrack-ng&lt;/code&gt;: not one tool but a suite, covering monitor mode setup,
packet capture, and the offline crack, all built around one weakness
in WPA2 -- the 4-way handshake that happens every time a client
associates.&lt;/p&gt;
&lt;p&gt;Everything below assumes you&#39;re testing a network you own or have
explicit written authorization to test -- a home lab AP, a CTF
range, or an engagement with signed scope. Capturing handshakes or
deauthing clients on a network that isn&#39;t yours is a computer crime
in most jurisdictions (CFAA in the US), full stop.&lt;/p&gt;
&lt;h2 id=&quot;what-you-need&quot;&gt;What you need&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;A wireless adapter that supports &lt;strong&gt;monitor mode&lt;/strong&gt; and &lt;strong&gt;packet
injection&lt;/strong&gt;. Most laptop-internal cards don&#39;t -- an external USB
adapter with a known-good chipset (Atheros AR9271, Realtek
RTL8812AU) is the usual recommendation.&lt;/li&gt;
&lt;li&gt;&lt;code&gt;aircrack-ng&lt;/code&gt; installed (&lt;code&gt;apt install aircrack-ng&lt;/code&gt; on Debian/Kali,
&lt;code&gt;pacman -S aircrack-ng&lt;/code&gt; on Arch).&lt;/li&gt;
&lt;li&gt;A wordlist. &lt;code&gt;rockyou.txt&lt;/code&gt; is the standard starting point -- fine
for proving a passphrase is &lt;em&gt;weak&lt;/em&gt;, useless against anything that
isn&#39;t.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;putting-the-interface-into-monitor-mode&quot;&gt;Putting the interface into monitor mode&lt;/h2&gt;
&lt;p&gt;NetworkManager and friends will fight you for the interface, so kill
them first:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;airmon-ng check kill&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;Then find the interface and switch it into monitor mode:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;iwconfig&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;airmon-ng start wlan0&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;This usually renames the interface to something like &lt;code&gt;wlan0mon&lt;/code&gt;.
Confirm with &lt;code&gt;iwconfig&lt;/code&gt; that it shows &lt;code&gt;Mode:Monitor&lt;/code&gt;.&lt;/p&gt;
&lt;h2 id=&quot;surveying-the-airspace&quot;&gt;Surveying the airspace&lt;/h2&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;airodump-ng wlan0mon&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;This lists every AP in range -- BSSID, channel, encryption, and the
clients (STATIONs) associated with each one. Find the target AP&#39;s
BSSID and channel and note them; the next command needs both.&lt;/p&gt;
&lt;h2 id=&quot;capturing-the-handshake&quot;&gt;Capturing the handshake&lt;/h2&gt;
&lt;p&gt;Lock onto the target and start writing packets to disk:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;airodump-ng -c &amp;lt;channel&amp;gt; --bssid &amp;lt;BSSID&amp;gt; -w capture wlan0mon&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;The goal is a WPA 4-way handshake, which fires whenever a client
associates with the AP. If a client is already connected, there&#39;s no
need to wait for a natural reconnect -- a deauth forces one on your
schedule instead of whenever someone happens to rejoin:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;aireplay-ng -0 5 -a &amp;lt;BSSID&amp;gt; -c &amp;lt;client MAC&amp;gt; wlan0mon&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;code&gt;-0 5&lt;/code&gt; sends 5 deauth packets. Watch the &lt;code&gt;airodump-ng&lt;/code&gt; window -- when
it shows &lt;code&gt;WPA handshake: &amp;lt;BSSID&amp;gt;&lt;/code&gt; in the top right corner, the
capture has what it needs. Stop it with Ctrl-C.&lt;/p&gt;
&lt;p&gt;An AP that supports &lt;strong&gt;PMKID&lt;/strong&gt; will hand over enough to attack without
any client interaction at all -- a known weakness in a lot of
consumer router firmware, and worth knowing about since it skips the
deauth step entirely. That&#39;s &lt;code&gt;hcxdumptool&lt;/code&gt; territory rather than the
airodump-ng/aireplay-ng flow above, but it&#39;s the quieter alternative
when one applies.&lt;/p&gt;
&lt;h2 id=&quot;cracking-it&quot;&gt;Cracking it&lt;/h2&gt;
&lt;p&gt;Point &lt;code&gt;aircrack-ng&lt;/code&gt; at the &lt;code&gt;.cap&lt;/code&gt; file and a wordlist:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;aircrack-ng -w /usr/share/wordlists/rockyou.txt -b &amp;lt;BSSID&amp;gt; capture-01.cap&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;If the passphrase is in the wordlist, this finds it. Under the hood
it&#39;s just hashing candidate passwords through PBKDF2 and comparing
against the handshake, so it&#39;s bottlenecked on CPU -- for anything
past a quick sanity check, &lt;code&gt;hashcat -m 22000&lt;/code&gt; on a GPU (after
converting the capture with &lt;code&gt;hcxpcapngtool&lt;/code&gt;) does the same attack an
order of magnitude faster.&lt;/p&gt;
&lt;h2 id=&quot;where-this-goes-wrong-in-practice&quot;&gt;Where this goes wrong in practice&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Wordlist mismatch.&lt;/strong&gt; &lt;code&gt;rockyou.txt&lt;/code&gt; catches short, common,
human-guessable passphrases and nothing else -- it says nothing
about a specific target the way a &lt;code&gt;crunch&lt;/code&gt;-generated or
target-aware list does (see the &lt;a href=&quot;/posts/crunch-hydra-wordlists&quot;&gt;Crunch/Hydra
post&lt;/a&gt; for building one).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No handshake, no attack.&lt;/strong&gt; Everything here depends on capturing
a full 4-way handshake or a PMKID. Miss it -- wrong channel, client
never reconnects, deauth packets dropped -- and there&#39;s nothing to
crack offline no matter how good the wordlist is.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;WPA3 breaks this entirely.&lt;/strong&gt; The handshake-capture-and-crack
model above is a WPA2 story. WPA3&#39;s SAE handshake is resistant to
offline dictionary attack by design -- if the AP under test
supports WPA3, turning it on is a better fix than a longer
passphrase.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Deauth has a blast radius.&lt;/strong&gt; &lt;code&gt;aireplay-ng -0&lt;/code&gt; knocks every
targeted client offline, not just the one you&#39;re trying to force a
handshake from. On a live network -- even one you&#39;re authorized to
test -- that&#39;s a disruption worth scoping and timing deliberately,
not firing off by habit.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Why You Can&#39;t Just Pi-Hack Your Neighbor&#39;s Bluetooth Speaker Quiet</title>
        <published>2026-08-26T00:00:00+00:00</published>
        <updated>2026-08-26T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/bluetooth-jamming-hijacking/"/>
        <id>https://legitbiz.xyz/posts/bluetooth-jamming-hijacking/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/bluetooth-jamming-hijacking/">&lt;p&gt;Search &quot;raspberry pi bluetooth jammer neighbor&quot; and you&#39;ll find no
shortage of forum threads from people who&#39;ve reached the same idea at
the same moment of frustration: a $35 board, a $10 USB dongle, and
surely there&#39;s a script that makes the noise stop. There isn&#39;t, not
one that works reliably, and more importantly, running one against a
device you don&#39;t own is illegal on its own regardless of whether it
works. Worth understanding why, and what the actual attack surface of
Bluetooth looks like, since the honest answer is more interesting
than the fantasy.&lt;/p&gt;
&lt;p&gt;This is a legal note as much as a technical one: interfering with
someone else&#39;s wireless device, jamming a frequency band, or
connecting to hardware you don&#39;t own without authorization violates
FCC regulations (in the US) and computer-intrusion or radio-
interference law in most other jurisdictions, independent of your
reason for doing it. A noise complaint is a civil/local matter; RF
jamming and unauthorized device access are federal-level problems.
None of what follows is a how-to against a real target -- it&#39;s why
the &quot;just take over their speaker&quot; idea doesn&#39;t hold up even before
the legal problem.&lt;/p&gt;
&lt;h2 id=&quot;what-a-bluetooth-speaker-is-actually-defended-by&quot;&gt;What a Bluetooth speaker is actually defended by&lt;/h2&gt;
&lt;p&gt;A paired, playing Bluetooth speaker isn&#39;t undefended, but its defenses
are mostly incidental rather than designed for this exact threat:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;It&#39;s already paired and connected.&lt;/strong&gt; Once a phone and speaker
complete pairing and are in an active streaming session, a third
device can&#39;t simply &quot;connect&quot; over the top of that -- Bluetooth
Classic (A2DP, the audio-streaming profile) is a point-to-point
link, and the speaker isn&#39;t listening for new connection requests
in the same way while it&#39;s already busy with one.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Frequency hopping.&lt;/strong&gt; Classic Bluetooth hops across 79 channels in
the 2.4GHz band roughly 1,600 times per second specifically to
resist both interference and eavesdropping. A naive jammer has to
cover the whole band continuously to have any effect, not just one
channel.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Range is short by design.&lt;/strong&gt; Class 2 Bluetooth (the class in
essentially every consumer speaker) is spec&#39;d for ~10 meters, and
real-world range through walls is often less. Anything built to
reach &quot;through my wall and into their speaker specifically&quot; is
already fighting physics the speaker&#39;s manufacturer didn&#39;t have to
think about.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;where-real-bluetooth-attacks-actually-live&quot;&gt;Where real Bluetooth attacks actually live&lt;/h2&gt;
&lt;p&gt;The security research that exists here is real, but it targets
specific implementation flaws, not &quot;any Bluetooth device, generically&quot; --
and understanding the difference is the whole point:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;BlueBorne (2017)&lt;/strong&gt; was a set of vulnerabilities in Bluetooth stack
implementations (Android, iOS, Windows, Linux) that allowed code
execution or connection takeover &lt;em&gt;without pairing&lt;/em&gt; -- but it worked
against specific unpatched stack versions, not Bluetooth as a
protocol, and was patched within weeks of disclosure across every
major OS. It&#39;s not a live technique against a current device.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;KNOB and BIAS attacks&lt;/strong&gt; exploited weaknesses in how some Bluetooth
implementations negotiated encryption key length and authentication,
letting an attacker downgrade or bypass parts of the handshake --
again, implementation-specific, disclosed to vendors, and patched.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Jamming (denial of service, not takeover)&lt;/strong&gt; is the crude option
people actually mean when they say &quot;Pi bluetooth jammer&quot; -- flooding
the 2.4GHz band with noise using an SDR or a modified Bluetooth
dongle in a tight loop. This doesn&#39;t hijack anything or play your
own audio; at best it makes the connection drop or stutter, at the
cost of also jamming your own WiFi, your neighbors&#39; WiFi (2.4GHz
WiFi shares the exact same band), any baby monitors, and anything
else nearby using that spectrum. It is also one of the most
straightforward RF violations to get caught for, because it&#39;s
trivially detectable with a $20 RF scanner and leaves an obvious,
continuous signature.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;BIAS-style social engineering (not a Bluetooth flaw at all)&lt;/strong&gt; is
actually the closest thing to what people picture -- a speaker left
in pairing mode, or one that accepts any new pairing request by
default, can genuinely be connected to by a stranger. This isn&#39;t a
hack, it&#39;s a device that was never taken out of an insecure default
state, and it depends entirely on the owner leaving it wide open,
not on anything a Pi script forces to happen.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;why-pi-dongle-specifically-doesn-t-get-you-there&quot;&gt;Why &quot;Pi + dongle&quot; specifically doesn&#39;t get you there&lt;/h2&gt;
&lt;p&gt;The Pi itself is a fine platform for Bluetooth experimentation -- BlueZ
(Linux&#39;s Bluetooth stack) runs on it natively, and tools like
&lt;code&gt;bluetoothctl&lt;/code&gt;, &lt;code&gt;hcitool&lt;/code&gt;, and &lt;code&gt;gattool&lt;/code&gt; are legitimate, widely used
software for working with Bluetooth Low Energy devices you own or
have permission to test. What doesn&#39;t exist is a script that turns
that into &quot;silence a specific speaker across a wall on demand&quot; --
between frequency hopping, short range, an already-active connection,
and every known implementation-level exploit being patched years ago,
the gap between &quot;forum post&quot; and &quot;working tool against a modern
device&quot; is the entire field of Bluetooth security research, not a
missing script.&lt;/p&gt;
&lt;h2 id=&quot;what-actually-solves-a-noisy-neighbor-problem&quot;&gt;What actually solves a noisy-neighbor problem&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Noise ordinance complaints&lt;/strong&gt; are usually time-of-day based (quiet
hours, typically 10pm-7am) and enforced by local police or code
enforcement -- documented, repeated complaints carry real weight and
don&#39;t require touching anyone&#39;s equipment.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A sound level app or a cheap dB meter&lt;/strong&gt; turns &quot;it&#39;s too loud&quot;
into a number a complaint or a landlord can act on, and a log with
timestamps is exactly what a noise ordinance case wants.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;HOA or landlord escalation&lt;/strong&gt; works faster than police in a lot of
shared-building situations, since lease violations are a lower bar
than a criminal complaint.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Passive mitigation on your own side&lt;/strong&gt; -- a white noise machine,
door/window seals, or a bookshelf against a shared wall -- is
unglamorous but is the only category of fix here that&#39;s both legal
and actually reliable, because it doesn&#39;t depend on the other
person&#39;s hardware cooperating with you.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Building Targeted Wordlists with Crunch, Then Feeding Them to Hydra</title>
        <published>2026-08-26T00:00:00+00:00</published>
        <updated>2026-08-26T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/crunch-hydra-wordlists/"/>
        <id>https://legitbiz.xyz/posts/crunch-hydra-wordlists/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/crunch-hydra-wordlists/">&lt;p&gt;Most password attacks don&#39;t win because of some clever exploit --
they win because the wordlist was better than the defense. A generic
list like &lt;code&gt;rockyou.txt&lt;/code&gt; is fine for catching lazy passwords, but it
says nothing about a specific target: a company&#39;s naming convention,
a device&#39;s default password pattern, a person&#39;s known habits. Crunch
exists to build the list that &lt;em&gt;does&lt;/em&gt; say something about the target,
and Hydra exists to burn through it against a live service. Neither
tool is exotic -- they&#39;re standard kit in any pentest lab, and they&#39;re
exactly as effective as the thinking that goes into the wordlist
before you run either one.&lt;/p&gt;
&lt;p&gt;Everything below assumes you&#39;re testing a system you own or have
explicit written authorization to test -- a home lab VM, a CTF box, or
an engagement with signed scope. Running this against anything else is
illegal in most jurisdictions, full stop.&lt;/p&gt;
&lt;h2 id=&quot;what-crunch-actually-does&quot;&gt;What Crunch actually does&lt;/h2&gt;
&lt;p&gt;Crunch generates every possible string that fits a pattern you give
it. It doesn&#39;t guess intelligently and it doesn&#39;t know anything about
the target -- it just enumerates a character space, which is exactly
what makes it predictable and scriptable. The basic form is:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch &amp;lt;min-len&amp;gt; &amp;lt;max-len&amp;gt; [charset] -o output.txt&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;min-len&lt;/code&gt; / &lt;code&gt;max-len&lt;/code&gt;&lt;/strong&gt; -- the shortest and longest password
length to generate. Setting both to the same number locks the
output to a fixed length, which matters a lot for output size (see
below).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;charset&lt;/code&gt;&lt;/strong&gt; -- the pool of characters to draw from. Omit it and
Crunch defaults to lowercase &lt;code&gt;a-z&lt;/code&gt;. You can hand it an explicit
string (&lt;code&gt;crunch 4 4 0123456789&lt;/code&gt;) or use one of the built-in sets in
&lt;code&gt;/usr/share/crunch/charset.lst&lt;/code&gt; (&lt;code&gt;-f charset.lst mixalpha-numeric&lt;/code&gt;,
for example).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;-o output.txt&lt;/code&gt;&lt;/strong&gt; -- write to a file instead of standard output.
For anything past a few thousand words, always write to a file --
see the piping section for why.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;A concrete example: every 4-digit numeric PIN --&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch 4 4 0123456789 -o pins.txt&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;That&#39;s 10,000 lines, instantly. Bump it to 6 digits and it&#39;s a
million lines. The output grows combinatorially with length and
charset size, so before generating anything, run the numbers with
&lt;code&gt;-c&lt;/code&gt; (line count estimate) or just &lt;code&gt;crunch &amp;lt;min&amp;gt; &amp;lt;max&amp;gt; &amp;lt;charset&amp;gt;&lt;/code&gt; with
no &lt;code&gt;-o&lt;/code&gt;, which prints the count and size before generating:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch 8 8 abcdefghijklmnopqrstuvwxyz&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;An 8-character all-lowercase list is already over 200 billion lines
and would eat terabytes of disk. This is the single most common way
people misuse Crunch -- setting a length and charset without checking
the size first, and filling a disk or waiting hours for a file they
didn&#39;t need.&lt;/p&gt;
&lt;h2 id=&quot;making-it-targeted-instead-of-exhaustive&quot;&gt;Making it targeted instead of exhaustive&lt;/h2&gt;
&lt;p&gt;Brute-forcing the entire keyspace is rarely the point -- most real
passwords follow a pattern, and Crunch&#39;s &lt;code&gt;-t&lt;/code&gt; flag lets you encode
that pattern directly instead of generating everything and filtering
after.&lt;/p&gt;
&lt;p&gt;&lt;code&gt;-t&lt;/code&gt; takes a template using placeholders:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code&gt;@&lt;/code&gt; -- lowercase letter&lt;/li&gt;
&lt;li&gt;&lt;code&gt;,&lt;/code&gt; -- uppercase letter&lt;/li&gt;
&lt;li&gt;&lt;code&gt;%&lt;/code&gt; -- digit&lt;/li&gt;
&lt;li&gt;&lt;code&gt;^&lt;/code&gt; -- symbol&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;So if recon on a target company shows their password policy is
&quot;capitalized word + two digits + symbol&quot; -- a pattern you&#39;d get from a
leaked credential, a policy document, or just observing one password
in the wild -- you encode it directly:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch 8 8 -t ,@@@@@%%^ -o targeted.txt&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;That collapses a search space that would otherwise be effectively
infinite down to something that finishes in minutes. This is the real
skill in using Crunch: the tool is dumb enumeration, so all the value
comes from constraining it with something you actually know about the
target -- a naming scheme, a known base word, a seasonal pattern
(&quot;Summer2026!&quot;-style corporate resets are still extremely common).&lt;/p&gt;
&lt;p&gt;You can also seed permutations of specific words instead of a full
character-space pattern, which is closer to what real password
attacks against humans look like:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch 6 10 -p password123 companyname admin&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;&lt;code&gt;-p&lt;/code&gt; (permute) generates every ordering of the given words instead of
a range of lengths, which is useful when you have a short list of
known-relevant terms (company name, product name, common defaults)
and want every combination rather than every possible string.&lt;/p&gt;
&lt;h2 id=&quot;piping-directly-into-hydra&quot;&gt;Piping directly into Hydra&lt;/h2&gt;
&lt;p&gt;Writing a multi-gigabyte wordlist to disk and then pointing Hydra at
the file works, but it&#39;s often unnecessary -- Crunch can stream
output straight into Hydra&#39;s stdin, which skips the disk write
entirely and starts the attack as soon as the first candidates exist
instead of waiting for generation to finish:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch 8 8 -t ,@@@@@%%^ | hydra -l admin -P - ssh://192.168.1.50&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;The pieces:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;-l admin&lt;/code&gt;&lt;/strong&gt; -- the known or assumed username. Use &lt;code&gt;-L users.txt&lt;/code&gt;
instead to test a list of usernames.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;-P -&lt;/code&gt;&lt;/strong&gt; -- tells Hydra to read the password list from stdin
(the &lt;code&gt;-&lt;/code&gt; is the placeholder for &quot;standard input&quot;) rather than a
file path.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;ssh://192.168.1.50&lt;/code&gt;&lt;/strong&gt; -- the target service and host. Hydra
supports dozens of protocols this way (&lt;code&gt;ftp://&lt;/code&gt;, &lt;code&gt;http-post-form://&lt;/code&gt;,
&lt;code&gt;rdp://&lt;/code&gt;, &lt;code&gt;smb://&lt;/code&gt;, etc.), each with its own syntax for the target
string.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;For a web login form instead of SSH, the target syntax gets more
specific, since Hydra needs to know the form fields and what a failed
attempt looks like:&lt;/p&gt;
&lt;pre class=&quot;giallo&quot; style=&quot;color: #CDD6F4; background-color: #1E1E2E;&quot; &gt;&lt;code data-lang=&quot;plain&quot;&gt;&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;crunch 6 8 -t @@@@@@%% | hydra -l admin -P - target.local http-post-form \&lt;/span&gt;&lt;/span&gt;
&lt;span class=&quot;giallo-l&quot;&gt;&lt;span&gt;  &amp;quot;/login:username=^USER^&amp;amp;password=^PASS^:Invalid credentials&amp;quot;&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;That last string has three colon-separated parts: the login path, the
POST body with &lt;code&gt;^USER^&lt;/code&gt;/&lt;code&gt;^PASS^&lt;/code&gt; as substitution markers, and a
string that appears in the response only on failure -- Hydra uses
that to distinguish a rejected attempt from a successful one.&lt;/p&gt;
&lt;h2 id=&quot;where-this-goes-wrong-in-practice&quot;&gt;Where this goes wrong in practice&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Throttling and lockouts.&lt;/strong&gt; Most real services rate-limit or
lock accounts after N failed attempts, which a raw Hydra run will
trip almost immediately. &lt;code&gt;-t&lt;/code&gt; on Hydra (unrelated to Crunch&#39;s &lt;code&gt;-t&lt;/code&gt;)
controls parallel connections -- turning it down (&lt;code&gt;-t 4&lt;/code&gt; instead of
the default 16) and adding &lt;code&gt;-W&lt;/code&gt; for a wait interval keeps an
engagement from just generating a wall of lockout alerts instead of
useful results.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Piping vs. pre-generating.&lt;/strong&gt; Streaming is convenient, but if
Hydra crashes or the connection drops mid-run, you lose your place
in the sequence with no way to resume -- Crunch doesn&#39;t checkpoint
a live stream. For long-running jobs, generating to a file (using
&lt;code&gt;crunch ... -o file.txt&lt;/code&gt; and optionally &lt;code&gt;split&lt;/code&gt; to chunk it) is
worth the disk space, since Hydra can resume against a file-backed
list.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Charset assumptions.&lt;/strong&gt; A list built purely from &lt;code&gt;charset.lst&lt;/code&gt;
guesses will almost always underperform a list built from the
target&#39;s own leaked data, breach corpora, or observed naming
conventions -- combining Crunch&#39;s pattern generation with
target-specific words (via &lt;code&gt;-p&lt;/code&gt;) consistently beats either approach
alone.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Verify scope before every run, not just once.&lt;/strong&gt; It&#39;s easy to
reuse a Hydra one-liner from a previous engagement and point it at
a new IP out of habit. The host in the command line is the entire
blast radius of a mistake here -- treat it with the same care as a
&lt;code&gt;rm -rf&lt;/code&gt; target.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>The MSR606: What a $30 Magstripe Reader/Writer Actually Does</title>
        <published>2026-08-26T00:00:00+00:00</published>
        <updated>2026-08-26T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/msr606-magstripe/"/>
        <id>https://legitbiz.xyz/posts/msr606-magstripe/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/msr606-magstripe/">&lt;p&gt;The MSR606 is a small USB device, widely sold on general electronics
marketplaces for well under $50, that reads and writes the magnetic
stripe on the back of a card. It shows up constantly in two very
different contexts: hobbyists and access-control installers using it
for exactly what it&#39;s built for, and card fraud, where it&#39;s the last
step that turns stolen card data into a physical, swipeable card. The
device itself is completely agnostic to which of those it&#39;s doing --
it just moves bytes onto a stripe. The legality and harm live entirely
in what data goes in and whose card it ends up on.&lt;/p&gt;
&lt;h2 id=&quot;what-s-actually-on-a-magnetic-stripe&quot;&gt;What&#39;s actually on a magnetic stripe&lt;/h2&gt;
&lt;p&gt;A card&#39;s stripe holds up to three parallel tracks of data, a format
standardized decades ago (ISO/IEC 7811) and still what nearly every
swipe-capable reader in the world expects:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Track 1&lt;/strong&gt; -- up to 79 alphanumeric characters, holding the
cardholder&#39;s name plus the same account number and expiration data
as Track 2. This is the track airline boarding-pass-style systems
and some access badges use, since it can carry a name and more
free-form data.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Track 2&lt;/strong&gt; -- up to 40 numeric characters, holding the primary
account number, expiration date, and service code. This is the
track that actually matters for a financial transaction -- most
payment terminals only read Track 2, and it&#39;s the track skimmers
are built specifically to capture.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Track 3&lt;/strong&gt; -- rarely used today, originally intended for
read-write applications like transit fare cards.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;None of this is encrypted or signed. A magstripe is a passive analog
recording -- whatever pattern of magnetic flux transitions is on the
tape gets read back as a fixed string of digits, with no
cryptographic check that the string came from a real, un-tampered
card. This is the entire reason chip cards (EMV) exist: a chip can
run a cryptographic challenge-response per transaction, while a
stripe just plays back the same static data every time, forever.
Anything that can read that static string can write it onto another
card.&lt;/p&gt;
&lt;h2 id=&quot;what-the-device-does-mechanically&quot;&gt;What the device does, mechanically&lt;/h2&gt;
&lt;p&gt;The MSR606 (and its more common sibling, the MSR605X) connects over
USB and ships with Windows software that talks to it over a simple
serial protocol. In practice, using it involves three separate
operations:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Read&lt;/strong&gt; -- swipe a card through the reader slot, and the software
decodes the raw flux transitions into the track strings and
displays them as text (the account number, expiration, and service
code on Track 2, for example).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Write&lt;/strong&gt; -- paste or type track data into the software, swipe a
blank magstripe card (sold in bulk, in the same aisle as the reader
itself, explicitly for this purpose -- hotel key stock, loyalty
card blanks, access badges), and the device re-encodes those exact
bytes onto the new card&#39;s stripe.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Erase&lt;/strong&gt; -- wipe a stripe back to blank, which is mostly useful for
reusing test cards during legitimate encoding work.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The coercivity setting (Hi-Co vs. Lo-Co) has to match the blank
card&#39;s stripe material or the write will fail to hold -- Hi-Co
stripes (the black ones, used on most bank cards) need a stronger
field than Lo-Co (the brown ones, common on hotel keys and gift
cards). This is the most common reason a first attempt at using one
of these devices doesn&#39;t work: wrong coercivity setting for the card
stock in hand.&lt;/p&gt;
&lt;h2 id=&quot;where-this-is-legitimate&quot;&gt;Where this is legitimate&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Access control and hotel systems.&lt;/strong&gt; Most hotel key encoders and a
fair number of building badge systems are, underneath the branded
housing, functionally identical hardware to a bare MSR605X --
installers and security integrators use standalone units like it
routinely to provision and test badges.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Point-of-sale and payment testing.&lt;/strong&gt; Developers and QA teams
building or certifying POS software need a way to generate known,
reproducible test card data without touching a real account -- an
encoder loaded with test-issuer data (the card networks publish
ranges specifically reserved for this) is standard lab equipment.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Loyalty and gift card programs.&lt;/strong&gt; Small businesses running their
own loyalty stripe cards use exactly this class of device to encode
new cards in-house instead of paying a vendor per batch.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Personal and hobbyist projects.&lt;/strong&gt; Encoding a personal library
card, an old transit card, or a DIY access system onto blank stock
you own is a long-running hobbyist niche with no card-network data
or another person&#39;s account involved anywhere in it.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The common thread: the data being written either belongs to the
person doing the writing, or was generated specifically as test data
that was never live in the first place.&lt;/p&gt;
&lt;h2 id=&quot;where-this-becomes-fraud&quot;&gt;Where this becomes fraud&lt;/h2&gt;
&lt;p&gt;The device doesn&#39;t change at all -- what changes is the source of the
Track 2 data being fed into the &quot;write&quot; step. Two supply chains feed
that step almost universally:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Skimmers.&lt;/strong&gt; A physical skimmer (a false front over an ATM or gas
pump card slot, or a compromised standalone reader) captures Track
1/2 data off real cards as they&#39;re swiped, often paired with a
hidden camera or a fake keypad overlay to grab the PIN separately.
That captured data gets pulled off the skimmer later and is now
identical in form to anything a legitimate operator would encode.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Carded data from breaches.&lt;/strong&gt; Track data harvested from
point-of-sale malware (memory-scraping malware reading card data
out of a terminal&#39;s RAM before it&#39;s encrypted -- the Target and
Home Depot breaches both worked this way) gets sold on card shops
in bulk, priced by issuing bank and country, exactly like the
account-number-only data sold for card-not-present fraud, but
formatted as full track dumps specifically because a buyer intends
to write it onto plastic and use it in person.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Once a buyer has Track 2 data from either source, an encoder and a
box of blank Hi-Co cards is the entire remaining toolchain to produce
a working physical clone -- no different mechanically from the
legitimate hotel-key use case above, just with someone else&#39;s account
number in the write buffer.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;EMV chip liability shift.&lt;/strong&gt; Since 2015 in the US (earlier
elsewhere), a merchant that accepts a chip card via magstripe swipe
instead of dipping the chip eats the fraud loss instead of the
issuing bank. This single policy change is the reason cloned-card
fraud at physical registers has dropped sharply compared to a
decade ago -- it didn&#39;t stop cloning, it made merchants stop
accepting the clone&#39;s swipe.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Chip cryptograms can&#39;t be cloned this way.&lt;/strong&gt; The chip generates a
unique cryptographic value per transaction; copying the static data
visible on the stripe (which most chip cards still carry, for
fallback compatibility) does not let an attacker reproduce that
value, so a cloned stripe from a chip card fails the moment it hits
a chip-reading terminal.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Tap-to-pay (NFC) has the same protection as chip&lt;/strong&gt;, using a
per-transaction cryptogram rather than replayable static data, and
is increasingly the default even at swipe-capable terminals.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Card-present anomaly detection.&lt;/strong&gt; Issuers flag a card used
in-person in a location inconsistent with its recent transaction
history or its cardholder&#39;s typical geography -- a cloned card
showing up hundreds of miles from the last legitimate swipe is a
strong, cheap signal independent of anything on the stripe itself.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Skimmer detection at the point of capture&lt;/strong&gt; is the actual
upstream fix: tamper-evident seals on ATMs and pumps, Bluetooth/RF
scanning for skimmer transmitters, and simply wiggling the card
reader before use, since most skimmers are shells that sit loosely
over the real slot.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>GitHub as a Recon Target: How Keys and Passwords End Up Public</title>
        <published>2026-08-15T00:00:00+00:00</published>
        <updated>2026-08-15T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/github-secret-recon/"/>
        <id>https://legitbiz.xyz/posts/github-secret-recon/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/github-secret-recon/">&lt;p&gt;Secrets end up in public GitHub repos constantly, and almost none of
it is the result of some elite hack. It&#39;s normal developer workflow,
at scale, with a &lt;code&gt;git push&lt;/code&gt; that happens a fraction of a second before
anyone remembers what was in the file. Finding those secrets
afterward isn&#39;t hard either -- it&#39;s one of the most automatable forms
of reconnaissance there is, which is exactly why it stays effective
year after year.&lt;/p&gt;
&lt;h2 id=&quot;how-the-secrets-get-there&quot;&gt;How the secrets get there&lt;/h2&gt;
&lt;p&gt;The same handful of mistakes account for most of it:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;.env&lt;/code&gt; files committed before &lt;code&gt;.gitignore&lt;/code&gt; catches up.&lt;/strong&gt; A new
project gets &lt;code&gt;git init&lt;/code&gt;&#39;d, a &lt;code&gt;.env&lt;/code&gt; with real database credentials
or API keys gets created for local dev, and the first commit --
before anyone&#39;s added &lt;code&gt;.env&lt;/code&gt; to &lt;code&gt;.gitignore&lt;/code&gt; -- ships it straight
into history.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&quot;Quick fix&quot; hardcoded credentials.&lt;/strong&gt; A key gets pasted directly
into a config file or a script to get something working under
deadline pressure, with every intention of moving it to an
environment variable later. Later doesn&#39;t always come before the
push does.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;CI/CD config files.&lt;/strong&gt; GitHub Actions workflows, Jenkinsfiles, and
Docker Compose files are magnets for this -- they&#39;re exactly the
place a database password or a cloud key needs to exist in plain
text for the pipeline to run, and they&#39;re routinely committed
alongside application code instead of pulled from a secrets
manager.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Secrets that survive their own deletion.&lt;/strong&gt; Removing a secret in a
later commit does not remove it from the repository -- it&#39;s still
sitting in the commit history, reachable by anyone who checks out
an earlier ref or runs &lt;code&gt;git log -p&lt;/code&gt;. A surprising number of &quot;we
fixed it&quot; commits are actually just adding a second copy of the
exposure.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Forks and mirrors.&lt;/strong&gt; If a secret was ever public, even briefly,
a fork or a clone made in that window can preserve it independently
of what the original repo does afterward -- deleting the original
repo, or making it private, doesn&#39;t touch copies that already exist
elsewhere.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;real-exposures-not-hypotheticals&quot;&gt;Real exposures, not hypotheticals&lt;/h2&gt;
&lt;p&gt;This isn&#39;t a theoretical risk. Toyota disclosed in 2022 that a GitHub
access key had been left in a public repository for almost five
years, granting access to a server holding data tied to roughly
296,000 customers -- the key was live the entire time, nobody was
looking. Mercedes-Benz had a similar incident surface in early 2024:
a GitHub token, exposed for about four years, that granted access to
the company&#39;s internal source code and infrastructure. In both cases
the failure wasn&#39;t sophisticated -- it was a credential in a public
repo that nobody noticed until someone went looking.&lt;/p&gt;
&lt;h2 id=&quot;why-it-s-so-easy-to-find&quot;&gt;Why it&#39;s so easy to find&lt;/h2&gt;
&lt;p&gt;GitHub&#39;s own code search indexes the full text of public repositories,
and it&#39;s usable by anyone with an account. Searching for a filename
like &lt;code&gt;.env&lt;/code&gt; or &lt;code&gt;credentials.json&lt;/code&gt; combined with a term like
&lt;code&gt;DB_PASSWORD&lt;/code&gt; or &lt;code&gt;SECRET_KEY&lt;/code&gt; turns up a meaningful hit rate on its
own, with zero tooling required -- this is often called &quot;GitHub
dorking,&quot; the same idea as a Google dork, applied to code search.&lt;/p&gt;
&lt;p&gt;Beyond manual searching, purpose-built scanners exist specifically to
automate this at scale. Tools like TruffleHog, Gitleaks, and Gitrob
work by pulling repositories (including their full commit history,
not just the current state) and running them through regex and
entropy-based detectors -- patterns that recognize the shape of an
AWS key (&lt;code&gt;AKIA&lt;/code&gt; followed by 16 characters), a private key block
(&lt;code&gt;-----BEGIN RSA PRIVATE KEY-----&lt;/code&gt;), or a Slack webhook URL, plus a
Shannon-entropy check that flags high-randomness strings likely to be
a token or password even without a recognizable prefix. These tools
were built for defenders auditing their own history, but the same
scan runs the same way against anyone else&#39;s public repo. Some of
them, along with dedicated projects like Shhgit, run continuously
against GitHub&#39;s public event stream, so a secret can be flagged
within seconds of the push that introduced it -- often faster than
the developer who committed it notices what they did.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Push protection.&lt;/strong&gt; GitHub&#39;s own secret scanning now blocks known
credential patterns (cloud provider keys, common API token formats)
at push time by default on public repos, rejecting the push before
the secret ever becomes visible -- the single most effective fix,
because it acts before exposure rather than after.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Pre-commit hooks.&lt;/strong&gt; Running a scanner like Gitleaks locally as a
pre-commit hook catches the mistake before it ever leaves the
developer&#39;s machine, which matters because once something is
pushed to a public remote, the exposure window is effectively
permanent regardless of what happens next.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Treat any pushed secret as compromised, immediately.&lt;/strong&gt; Deleting
the file, force-pushing over the commit, or rewriting history does
not undo the exposure -- if it was public even briefly, assume it
was seen, and rotate the credential rather than just removing it.
History rewrites clean up the repo; they don&#39;t un-ring the bell.
Automated scanners are frequently faster than the response time of
the person who made the mistake.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Short-lived, scoped credentials.&lt;/strong&gt; A key that expires in hours
instead of never, and that can only touch one resource instead of
an entire account, turns a catastrophic leak into a minor one. This
is the same logic as least-privilege access anywhere else -- assume
leakage will happen occasionally, and design the blast radius down.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Secrets managers over config files, as a default habit.&lt;/strong&gt; Vault,
AWS Secrets Manager, and equivalents exist specifically so a
credential never has to live in a file that could plausibly end up
in a commit. The fix that scales isn&#39;t &quot;be more careful&quot; -- it&#39;s
removing the opportunity for the mistake to happen at all.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>The Idiot&#39;s Guide to Fraud, Part 1: The Fraud Triangle</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/fraud-triangle/"/>
        <id>https://legitbiz.xyz/posts/fraud-triangle/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/fraud-triangle/">&lt;p&gt;Welcome to the first installment of a series nobody asked for and HR
would very much like me to stop writing. We&#39;re talking about fraud --
the theory of it, the mechanics of it, the glorious, avoidable
stupidity of it -- across banking, corporate accounting, and organized
retail theft. I promise you an unreasonable amount of detail. Where
that detail comes from is between me and my lawyer, and no, that&#39;s not
a bit, that&#39;s just a boundary.&lt;/p&gt;
&lt;p&gt;Let&#39;s start at the beginning, with the closest thing this field has to
a periodic table: the &lt;strong&gt;Fraud Triangle&lt;/strong&gt;. Every scheme you will ever
read about, no matter how baroque, boils down to three ingredients
sitting in the same room at the same time. Take any one of them away
and the whole thing quietly fails to happen. This is, coincidentally,
also why &quot;just hire honest people&quot; has never once worked as a fraud
prevention strategy -- honesty is not a fixed trait, it&#39;s a variable
that moves depending on how the other two legs of the triangle are
doing that week.&lt;/p&gt;
&lt;p&gt;The concept comes from Donald Cressey, a criminologist who in the
1950s did the very sensible thing of just asking convicted embezzlers
why they did it, instead of guessing. Turns out people who steal from
their employers are, shockingly, willing to explain their reasoning in
great detail once there&#39;s no more upside to keeping quiet. Somebody
should try that approach more often.&lt;/p&gt;
&lt;h2 id=&quot;pressure-the-non-shareable-problem&quot;&gt;Pressure: the &quot;non-shareable problem&quot;&lt;/h2&gt;
&lt;p&gt;Every fraud starts with a problem the person doesn&#39;t feel they can
say out loud. Gambling losses. A second mortgage nobody at the office
knows about. A lifestyle that quietly outgrew the salary funding it. A
bonus target invented by someone in a much nicer office who will never
have to explain how it gets hit. The word doing all the work here is
&lt;em&gt;non-shareable&lt;/em&gt; -- not &quot;the person has a problem,&quot; but &quot;the person has
a problem they&#39;ve decided they&#39;d rather commit a felony over than
mention to their manager, their spouse, or a financial advisor.&quot;&lt;/p&gt;
&lt;p&gt;This is why &quot;our people would tell us if something was wrong&quot; is one
of the great punchlines of corporate risk management. They wouldn&#39;t.
That&#39;s the entire premise. The problem is non-shareable specifically
&lt;em&gt;to the people who could actually help.&lt;/em&gt; A functioning EAP program
catches more fraud, dollar for dollar, than most compliance training
ever will, and almost nobody budgets for it that way.&lt;/p&gt;
&lt;p&gt;Pressure is also the one leg of the triangle you, the organization,
mostly can&#39;t see coming and mostly can&#39;t control. You don&#39;t get to
audit somebody&#39;s marriage or their sports book account. Which is
exactly why the smart money -- and by &quot;smart money&quot; I mean people who
have thought about this far more carefully than a rational person
should -- focuses defensive effort on the other two legs instead.&lt;/p&gt;
&lt;h2 id=&quot;opportunity-the-only-leg-you-actually-control&quot;&gt;Opportunity: the only leg you actually control&lt;/h2&gt;
&lt;p&gt;This is the fun one. Opportunity is the belief -- correct or not -- that
the scheme can be executed and concealed. And unlike pressure and
rationalization, which live entirely inside somebody&#39;s skull where you
have no jurisdiction, opportunity lives in your process design, which
you absolutely do control and routinely leave lying around like an
unlocked bike.&lt;/p&gt;
&lt;p&gt;Opportunity shows up as:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;One person with both custody and recordkeeping.&lt;/strong&gt; The same
person who receives the cash also reconciles the cash. The same
person who approves the vendor also cuts the vendor&#39;s check. Anyone
who has ever taken an intro auditing class will tell you this is the
cardinal sin of internal controls, and yet it is &lt;em&gt;everywhere&lt;/em&gt;,
because segregation of duties costs headcount and headcount costs
budget and budget is somebody else&#39;s problem this quarter.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Reconciliation theater.&lt;/strong&gt; A control that exists on an org chart
and a compliance binder but that nobody has actually performed with
their eyes open in eleven months. Deeply comforting to auditors who
only check that the checkbox exists, not that anyone meant it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Trust standing in for verification.&lt;/strong&gt; &quot;Dave&#39;s been here
nineteen years, he&#39;d never.&quot; Dave has been here nineteen years
specifically because nobody has ever checked Dave&#39;s work, and Dave
noticed that around year three.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Override authority with no counterweight.&lt;/strong&gt; Somebody senior
enough to approve their own exceptions. This one shows up constantly
once we get to the accounting fraud posts -- it&#39;s basically the
entire Enron plot in one bullet point.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;None of this requires the fraudster to be a criminal mastermind. Most
aren&#39;t. Most fraud is not &lt;em&gt;Ocean&#39;s Eleven&lt;/em&gt;, it&#39;s a mediocre plan that
worked purely because nobody was checking. This is either deeply
reassuring or deeply terrifying depending on whether you&#39;re reading
this as a business owner or, hypothetically, someone with a more
colorful employment history.&lt;/p&gt;
&lt;p&gt;Case in point, and a genuine gift to anyone teaching this concept:
Thomas Frudaker, arrested in Yuma, Arizona in 2018 after a store
employee noticed something off about a computer he&#39;d just returned.
&quot;Off&quot; turned out to mean gutted -- the valuable internal components
quietly removed, the shell reassembled to look untouched, then handed
back across the counter for a full refund. Investigators believe he&#39;d
been running the identical play at Walmart stores nationwide for about
eighteen months, an estimated &lt;strong&gt;thousand-plus locations&lt;/strong&gt;, netting
something like &lt;strong&gt;$1.3 million&lt;/strong&gt; before one employee at one register
finally opened the box. Not a hacking operation, not a criminal
syndicate -- one guy, a screwdriver, and a return desk that verified
&quot;is there a computer-shaped object in this box&quot; rather than &quot;is this
the computer that was sold.&quot; That gap between the two is opportunity,
in its purest and most poorly-defended form. Eighteen months is a long
time to profit off a policy nobody thought to close.&lt;/p&gt;
&lt;p&gt;I bring this up not to admire the technique -- perish the thought -- but
because it&#39;s the cleanest illustration you&#39;ll find of why &quot;opportunity&quot;
gets top billing among the three legs. Pressure and rationalization
are guesswork. Opportunity is a policy decision, and this particular
policy decision cost a trillion-dollar retailer over a million dollars
because verifying a return took longer than trusting it.&lt;/p&gt;
&lt;h2 id=&quot;rationalization-the-bedtime-story-you-tell-yourself&quot;&gt;Rationalization: the bedtime story you tell yourself&lt;/h2&gt;
&lt;p&gt;The last ingredient is the one people underrate, because it sounds
like a soft, feelings-based footnote next to &quot;stole $400,000.&quot; It
isn&#39;t. Rationalization is the mental permission slip that lets someone
commit fraud and still show up the next morning believing they&#39;re
fundamentally a good person. Nobody -- and I do mean nobody -- sits down
and thinks &quot;I am now going to be a thief.&quot; They think:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&quot;I&#39;m borrowing it. I&#39;ll put it back before anyone checks.&quot;&lt;/li&gt;
&lt;li&gt;&quot;They underpay me for what I actually do here. I&#39;m owed this.&quot;&lt;/li&gt;
&lt;li&gt;&quot;Everyone in this industry does some version of this, I&#39;m just
the one being honest about it, in the sense of doing it and not
discussing it.&quot;&lt;/li&gt;
&lt;li&gt;&quot;The company makes plenty. This isn&#39;t going to hurt anyone.&quot;&lt;/li&gt;
&lt;li&gt;&quot;I&#39;m smarter than the people whose job it is to catch this,
and being right about that isn&#39;t a crime.&quot;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;That last one is doing a lot of quiet, structural work, by the way.
Confidence in one&#39;s own cleverness is practically load-bearing in this
field -- ask literally anyone who&#39;s spent time around people who do
this for a living. Not that I have. Allegedly.&lt;/p&gt;
&lt;p&gt;Rationalization matters because it&#39;s the difference between fraud and
ordinary theft, psychologically speaking. A mugger doesn&#39;t need a
philosophy. A fraudster generally does, because they have to look
themselves in the mirror, go to their kid&#39;s soccer game, and sit
through Thanksgiving dinner. The story doesn&#39;t have to be &lt;em&gt;good.&lt;/em&gt; It
just has to be good enough to get them through the next reflection in
the bathroom mirror.&lt;/p&gt;
&lt;h2 id=&quot;why-this-triangle-is-the-whole-syllabus&quot;&gt;Why this triangle is the whole syllabus&lt;/h2&gt;
&lt;p&gt;Here&#39;s the practical payoff, and the reason this isn&#39;t just true-crime
trivia dressed up as theory: whenever you look at any fraud scheme --
a bank teller running fake accounts, a CFO booking revenue that
doesn&#39;t exist yet, a booster ring fencing stolen retail goods through
a marketplace storefront -- and you ask &quot;what was the pressure, what
was the opportunity, what was the rationalization,&quot; the entire
mechanism falls out cleanly, every single time. It&#39;s a diagnostic, not
a decoration.&lt;/p&gt;
&lt;p&gt;It also tells you exactly where your money is best spent defensively,
which is the part actual fraud examiners get paid to know and actual
executives routinely ignore in favor of a training video nobody
watches past minute two. You cannot screen job applicants for future
gambling debt. You cannot audit someone&#39;s internal monologue for
rationalization quality. But you can absolutely, unglamorously, close
down opportunity -- segregate duties, verify instead of trust,
reconcile for real, remove unilateral override authority -- and that
one leg of the triangle does more preventive work than the other two
combined. This is not a hot take. This is close to the entire field of
internal controls in one sentence, and consultants will bill you
five figures to tell you the same thing with a nicer slide deck.&lt;/p&gt;
&lt;p&gt;One more wrinkle before we move on, because the original three-legged
model has taken some deserved criticism over the decades: later
researchers proposed a fourth leg, capability -- the idea that pulling
off a &lt;em&gt;sophisticated&lt;/em&gt; fraud (as opposed to a clumsy one that gets
caught in a week) takes more than motive and opportunity. It takes the
actual skill to design the scheme, the composure to execute it under
pressure, and the social confidence to talk your way past questions
when someone finally gets suspicious. This became the &quot;Fraud Diamond,&quot;
and it&#39;s worth filing away now, because it&#39;s going to matter a great
deal once we get into the accounting-fraud posts, where the people
involved were rarely improvising.&lt;/p&gt;
&lt;p&gt;Next up: bank fraud, starting with account takeover and synthetic
identity schemes -- which, despite what the industry&#39;s marketing
materials imply, do not primarily get stopped by better passwords.&lt;/p&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Hello, World</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/hello-world/"/>
        <id>https://legitbiz.xyz/posts/hello-world/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/hello-world/">&lt;p&gt;First post. This blog is a Zola site, hand-rolled terminal theme, built
and deployed from this box straight to Apache&#39;s docroot. No JS, no
database, no CMS -- just markdown files and &lt;code&gt;zola build&lt;/code&gt;.&lt;/p&gt;
&lt;p&gt;More to come as there&#39;s more to write about.&lt;/p&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>RDP Shops and the Geography Problem in Card Fraud</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/rdp-and-card-fraud/"/>
        <id>https://legitbiz.xyz/posts/rdp-and-card-fraud/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/rdp-and-card-fraud/">&lt;p&gt;Card networks and processors lean hard on geography as a fraud
signal: does the IP address, the billing address, and the issuing
bank&#39;s country all roughly agree? A mismatch is one of the cheapest,
highest-signal checks in the whole risk stack. Which is why a
specific piece of infrastructure -- the compromised RDP box -- has
stayed valuable to card fraud for over a decade, long after most of
the rest of the tooling around it has changed.&lt;/p&gt;
&lt;h2 id=&quot;why-rdp-specifically&quot;&gt;Why RDP specifically&lt;/h2&gt;
&lt;p&gt;A stolen card has an issuing bank, and that bank has a country, often
down to a metro area. A transaction placed from an IP in that same
metro area looks nothing like one placed from a datacenter halfway
around the world. Renting -- or more often, buying access to an
already-compromised -- residential or small-business Windows machine
in the right city gives a fraudster a local IP, a local system clock,
and a local browser fingerprint for the price of a remote desktop
session.&lt;/p&gt;
&lt;p&gt;This is different from a residential proxy, which routes traffic
through someone&#39;s connection but still runs the fraudster&#39;s own
browser and OS. RDP hands over an entire real machine: real installed
fonts, a real hardware ID, a browser profile with actual history in
it. Fingerprinting systems built to catch proxies frequently wave
this straight through, because by every signal they check, it &lt;em&gt;is&lt;/em&gt; a
real consumer machine -- it just has two people using it.&lt;/p&gt;
&lt;h2 id=&quot;how-the-access-gets-sold&quot;&gt;How the access gets sold&lt;/h2&gt;
&lt;p&gt;RDP access is commoditized on dedicated marketplaces, priced by
country, city, and sometimes by the specific bank whose cardholders
live nearby. Machines are sourced from mass credential-stuffing
against exposed RDP ports (3389 left open to the internet is still
astonishingly common), and turnover is high -- a box gets used for
card testing until its IP is burned, then abandoned for the next one.&lt;/p&gt;
&lt;p&gt;Card testing itself -- running small, fast transactions across many
stolen card numbers to find which ones are still live -- is usually
the first thing to happen on a freshly acquired box, before it gets
used for anything larger. That testing traffic has its own shape:
short session, narrow range of merchant categories, high transaction
velocity, low individual amounts.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Flag datacenter and known-RDP-provider ASNs outright&lt;/strong&gt;, and treat
&quot;clean&quot; residential ASNs with sudden unfamiliar login patterns
(new device, new user-agent, odd hours) as elevated risk rather
than trusted by default.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Look past IP geography to session behavior.&lt;/strong&gt; RDP sessions often
have telltale latency and input-timing signatures -- mouse movement
and keystroke timing over a remote session differs measurably from
local input, even when every other fingerprint matches.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Rate-limit and pattern-match card testing specifically&lt;/strong&gt;: many
small authorizations across many card numbers in a short window,
from one session or device, is a distinct signature from normal
shopping behavior and is cheap to detect independent of geography.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;3-D Secure (3DS2)&lt;/strong&gt; shifts liability and adds a real-time bank-side
challenge that a hijacked RDP session usually can&#39;t complete,
since it doesn&#39;t grant access to the cardholder&#39;s phone or bank
app.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Close exposed RDP ports.&lt;/strong&gt; The supply side of this entire market
is machines with 3389 open to the internet and weak or reused
credentials. This is the cheapest fix in the chain and the one most
consistently skipped.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>The COD Handoff Gap</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/the-cod-handoff-gap/"/>
        <id>https://legitbiz.xyz/posts/the-cod-handoff-gap/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/the-cod-handoff-gap/">&lt;p&gt;A case study in what happens when two systems each assume the other
one is tracking payment state, and neither one actually is. This
described a real class of bug in food-delivery integrations; the
specific gap here has since been closed by tightening how orders get
handed off to third-party couriers.&lt;/p&gt;
&lt;h2 id=&quot;the-setup&quot;&gt;The setup&lt;/h2&gt;
&lt;p&gt;Plenty of restaurants run delivery in-house during normal hours and
outsource to a gig courier platform (DoorDash, Uber Eats, etc.) after
their own drivers go home. The restaurant&#39;s own ordering site
supported cash-on-delivery as a payment option -- a reasonable thing
to offer when the restaurant&#39;s own driver collects payment at the
door.&lt;/p&gt;
&lt;h2 id=&quot;the-gap&quot;&gt;The gap&lt;/h2&gt;
&lt;p&gt;The order-routing logic didn&#39;t distinguish payment method when
deciding whether to outsource a late order to a courier platform. An
order marked &quot;cash on delivery&quot; got dispatched to the courier the
same way a prepaid order would. The courier platform&#39;s side of the
integration had no concept of cash orders at all -- its whole model
assumes the platform processed payment, so it treated every
handed-off order as already paid, and the courier had no prompt to
collect anything at the door.&lt;/p&gt;
&lt;p&gt;The result: an order placed after normal delivery hours, marked cash,
routed automatically to a courier who believed (correctly, from their
side of the integration) that payment was already settled. Nobody in
the chain was lying to their own system -- both systems were correct
about their own state and wrong about each other&#39;s.&lt;/p&gt;
&lt;h2 id=&quot;why-this-shape-of-bug-is-common&quot;&gt;Why this shape of bug is common&lt;/h2&gt;
&lt;p&gt;This is a classic boundary failure between two systems that don&#39;t
share a payment-state field. The restaurant&#39;s POS knows &quot;unpaid,
collect cash.&quot; The courier dispatch API has no field for that,
because it was designed around a single assumption: if an order
reaches the courier, Stripe or the platform&#39;s own processor already
settled it. Neither system was broken in isolation. The bug only
existed in the handoff, which is exactly why it&#39;s easy to miss in
testing -- most QA exercises each system against its own happy path,
not the seam between two vendors&#39; assumptions about each other.&lt;/p&gt;
&lt;h2 id=&quot;where-it-got-fixed&quot;&gt;Where it got fixed&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Payment status became a required field in the dispatch payload&lt;/strong&gt;,
not an assumed constant. A courier API that can&#39;t represent &quot;unpaid,
collect on delivery&quot; will eventually be handed one anyway.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;COD orders got excluded from auto-outsourcing&lt;/strong&gt; rather than
patched to carry payment state through -- the simpler fix, and the
one most integrations of this kind land on, since teaching a
third-party courier platform to collect cash reliably is a bigger
lift than just not outsourcing those orders.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Order routing logic got a payment-method check before the
after-hours outsourcing decision&lt;/strong&gt;, closing the specific timing
window (after in-house delivery cutoff, before the courier
integration&#39;s own hours logic kicked in) where the mismatch used to
slip through.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The general lesson outlives this specific fix: any time an order,
payment, or identity handoff crosses a vendor boundary, check what
each side assumes the other one already verified. That assumption gap
is where this entire category of bug lives.&lt;/p&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>The Fraud Score</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/the-fraud-score/"/>
        <id>https://legitbiz.xyz/posts/the-fraud-score/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/the-fraud-score/">&lt;p&gt;Almost every online transaction gets a number attached to it before
it&#39;s approved, and almost no consumer ever sees it. Processors call
it a risk score, a trust score, a fraud score depending on the
vendor -- Stripe Radar, Sift, card-network tools like Visa Advanced
Authorization all produce some version of the same thing: a
0-to-100-ish estimate of how likely this specific transaction is to
get charged back.&lt;/p&gt;
&lt;h2 id=&quot;what-actually-feeds-it&quot;&gt;What actually feeds it&lt;/h2&gt;
&lt;p&gt;The score isn&#39;t one signal, it&#39;s a blend, and the blend is the whole
value of the product:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Identity age and stability&lt;/strong&gt; -- how long has this email, phone
number, device, or card been seen before, and has it been
consistent. A card first seen ten minutes ago, paired with an email
created yesterday, is a very different prior than a card with two
years of clean history.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Velocity&lt;/strong&gt; -- how many transactions, on how many cards, from this
device or IP or shipping address, in how short a window. Velocity
across seemingly unrelated accounts is one of the strongest signals
there is, because it&#39;s expensive for a fraudster to fully
compartmentalize every session.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Network reputation&lt;/strong&gt; -- has this IP, device fingerprint, or card
BIN shown up attached to confirmed fraud anywhere else in the
processor&#39;s cross-customer data. This is why a single processor
like Stripe or a card network can catch patterns an individual
merchant never could: they see fraud reported across thousands of
unrelated businesses.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Behavioral and environmental noise&lt;/strong&gt; -- mismatched timezone vs.
billing address, browser languages that don&#39;t match geography,
autofill patterns that look scripted rather than typed, checkout
speed that&#39;s inhumanly fast.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Each signal alone is weak and full of false positives. A VPN user, a
shared family card, someone shopping while traveling -- all trip
individual flags constantly. The score&#39;s entire job is combining
weak, noisy signals into something with better precision than any one
of them.&lt;/p&gt;
&lt;h2 id=&quot;how-it-gets-used&quot;&gt;How it gets used&lt;/h2&gt;
&lt;p&gt;Most systems reduce the score to three buckets: allow, block, or
send to manual/step-up review (OTP, 3DS challenge, human look).
Merchants tune the thresholds against their own risk tolerance --
tighter thresholds cut fraud losses but also decline legitimate
customers, and that false-decline cost is often larger than the fraud
it prevents, which is why thresholds are a business decision as much
as a security one.&lt;/p&gt;
&lt;h2 id=&quot;where-fraud-tries-to-hide-under-the-line&quot;&gt;Where fraud tries to hide under the line&lt;/h2&gt;
&lt;p&gt;None of this is secret from the people trying to beat it, which is
why the more sophisticated fraud rings optimize for staying under
the threshold rather than avoiding detection outright: smaller
transaction amounts, warmed-up accounts aged for weeks before use,
one card per session instead of testing many, deliberately slow and
human-paced checkout flows. This is the actual arms race -- not
&quot;detect fraud&quot; vs. &quot;hide from detection,&quot; but a threshold-tuning
contest on both sides.&lt;/p&gt;
&lt;p&gt;The practical takeaway for anyone building or buying this kind of
system: a fraud score is a probability estimate, not a verdict, and
treating it as binary (block everything under 40, allow everything
above) throws away the information in &lt;em&gt;why&lt;/em&gt; a transaction scored the
way it did. The reason code behind the score is usually more useful
for a human reviewer than the number itself.&lt;/p&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>Triangulation Fraud</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/triangulation-fraud/"/>
        <id>https://legitbiz.xyz/posts/triangulation-fraud/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/triangulation-fraud/">&lt;p&gt;A pattern where the seller never owns any inventory, and the buyer
often never realizes anything was wrong -- the item shows up, on
time, as described. The fraud is invisible to the one person it looks
like it should hurt most.&lt;/p&gt;
&lt;h2 id=&quot;the-mechanics&quot;&gt;The mechanics&lt;/h2&gt;
&lt;p&gt;A seller lists a product on Marketplace A at a normal or slightly
below-market price and takes a legitimate payment from a real buyer.
Instead of shipping anything they own, the seller turns around and
buys the same item from Marketplace B or a retailer&#39;s own site, using
a stolen card, and enters the real buyer&#39;s address as the shipping
destination. The item ships direct from Marketplace B to the buyer.
The buyer gets exactly what they ordered, on a normal timeline, and
has no reason to suspect anything.&lt;/p&gt;
&lt;p&gt;The seller&#39;s payout from Marketplace A is now pure profit -- they
spent nothing, since the stolen card covered the actual purchase.
The loss lands entirely on Marketplace B (or its payment processor)
and the cardholder whose card was used without authorization, once
the chargeback comes in weeks later. By the time it does, the seller
has moved on to the next order, and the connection between &quot;stolen
card used on Marketplace B&quot; and &quot;seller account on Marketplace A&quot; is
not something either marketplace can see on their own -- they&#39;re
different companies with no shared view of the transaction.&lt;/p&gt;
&lt;h2 id=&quot;why-it-works&quot;&gt;Why it works&lt;/h2&gt;
&lt;p&gt;The scheme survives specifically because it splits a single fraud
across two unrelated systems, each of which only sees half the
picture. Marketplace A sees a normal completed sale with a happy
buyer and no chargeback on their books -- there&#39;s nothing for their
fraud systems to flag. Marketplace B sees an order paid with a stolen
card shipped to a third-party address, which does look like fraud to
them individually, but by the time the chargeback resolves, it&#39;s
processed as an isolated stolen-card incident, not connected back to
a seller pattern on a completely different platform.&lt;/p&gt;
&lt;p&gt;The seller&#39;s real exposure is entirely on the Marketplace B side, and
even that takes time to catch up with them, since a chargeback
investigation is slower than the sell-ship-profit cycle it&#39;s chasing.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Billing-address / shipping-address mismatch, treated as high risk
rather than routine.&lt;/strong&gt; A huge share of legitimate e-commerce ships
to a different address than billing (gifts, offices), so this alone
is weak -- but combined with a brand-new account or a shipping
address that&#39;s never been used on that card before, it&#39;s one of the
strongest available signals.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Cross-marketplace pattern matching.&lt;/strong&gt; Fraud-prevention vendors
(Forter, Riskified, Signifyd) sit across many merchants at once
specifically to catch this seam -- a card used fraudulently on
Marketplace B and a seller account with unusual fulfillment timing
on Marketplace A are invisible to each platform alone, but visible
to a network that sees both.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Fulfillment-source verification.&lt;/strong&gt; Marketplaces that require
sellers to prove they hold inventory, or that flag sellers whose
shipments consistently originate from a different retailer&#39;s
packaging or return address, catch this at the seller-account level
rather than the transaction level.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Velocity on new seller accounts.&lt;/strong&gt; A brand-new seller account
doing high-value, high-volume, drop-ship-pattern sales in its first
days is a strong prior on its own, independent of any single
transaction looking clean.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Faster chargeback signal loops.&lt;/strong&gt; The core weakness of the scheme
is the lag between &quot;card gets used&quot; and &quot;chargeback lands and gets
attributed.&quot; Shortening that loop, and sharing the attribution back
to the originating seller account rather than treating it as an
isolated card incident, is what actually closes the gap -- not any
single transaction-level check.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
    <entry xml:lang="en">
        <title>The Wipe-and-Flip Pipeline</title>
        <published>2026-08-14T00:00:00+00:00</published>
        <updated>2026-08-14T00:00:00+00:00</updated>
        
        <author>
          <name>Unknown</name>
        </author>
        
        <link rel="alternate" type="text/html" href="https://legitbiz.xyz/posts/wipe-and-flip-pipeline/"/>
        <id>https://legitbiz.xyz/posts/wipe-and-flip-pipeline/</id>
        
        <content type="html" xml:base="https://legitbiz.xyz/posts/wipe-and-flip-pipeline/">&lt;p&gt;Notes on a fraud pattern that&#39;s common enough to have a shape: a
laptop gets stolen, wiped, and relisted as clean secondhand hardware,
often within 48 hours. The interesting part isn&#39;t the theft -- it&#39;s
where the pipeline actually breaks.&lt;/p&gt;
&lt;h2 id=&quot;the-take&quot;&gt;The take&lt;/h2&gt;
&lt;p&gt;Most stolen laptops don&#39;t come from break-ins. They come from
unattended bags -- a coffee shop table, a rideshare back seat, a
conference badge line. Opportunistic theft outnumbers targeted
burglary by a wide margin because it scales: one person working a
busy cafe for an afternoon walks away with more inventory than a
night spent on a residential break-in, for a fraction of the risk.&lt;/p&gt;
&lt;p&gt;The second-largest source is quieter: assets that never get returned.
An employee leaves a company, a rental laptop doesn&#39;t come back, a
&quot;lost&quot; device gets reported once and never chased. Orgs without real
asset tracking absorb this as background loss, which is exactly what
makes it attractive to insiders.&lt;/p&gt;
&lt;h2 id=&quot;the-wipe&quot;&gt;The wipe&lt;/h2&gt;
&lt;p&gt;This stage decides whether the device is worth anything. A laptop
with an active account-level lock -- tied to the owner&#39;s cloud
account rather than to the hardware -- is close to worthless on the
resale market, because the lock survives a factory reset, an OS
reinstall, and usually a full drive swap.&lt;/p&gt;
&lt;p&gt;The lock isn&#39;t on the drive. That&#39;s the detail most attempts get
wrong, and it&#39;s why wipe-and-flip operations increasingly skip recent
flagship laptops and focus on older or enterprise-fleet hardware
where the lock was never enabled, was tied to an MDM system that got
deprovisioned wrong, or never existed. Asset tags and BIOS labels get
removed or sanded off; serials get left out of listings entirely.&lt;/p&gt;
&lt;h2 id=&quot;the-channel&quot;&gt;The channel&lt;/h2&gt;
&lt;p&gt;Wiped inventory moves through three channels, in order of preference:
peer-to-peer marketplaces with no ID or serial verification, pawn
shops in jurisdictions with weak reporting requirements, and bulk
wholesale export -- pallets of &quot;used electronics, untested&quot; sold by
weight to overseas resellers who have no practical way to check a
serial against a theft report even if they wanted to.&lt;/p&gt;
&lt;p&gt;Cash-only, local-pickup listings are the default, since they skip
payment processor identity checks and shipping labels -- both of
which leave a paper trail. Listings often describe a fully functional
device as &quot;for parts,&quot; a small hedge that lets the seller claim
ignorance later.&lt;/p&gt;
&lt;h2 id=&quot;where-it-breaks&quot;&gt;Where it breaks&lt;/h2&gt;
&lt;p&gt;Two structural weak points, and defenses aimed at them do more work
than anything aimed at preventing the theft itself:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Account-bound remote lock, enabled by default.&lt;/strong&gt; Find My /
Activation Lock on Apple hardware, Find Hub on Chromebooks, and
Windows Find My Device all survive a wipe if they were turned on
before the theft. The gap is almost always &quot;never enabled,&quot; not
&quot;defeated.&quot;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Firmware-level persistence agents.&lt;/strong&gt; Tools like Absolute
Persistence live in UEFI/BIOS, not the OS, and reinstall themselves
after a factory reset or drive swap -- pinging the owner the moment
the device touches the internet again.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Serial / IMEI blacklist registries.&lt;/strong&gt; CTIA/GSMA&#39;s device registry
and similar theft-report databases let a buyer check a serial
before paying. The channel only works because most secondhand
buyers never check.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Real offboarding for org-owned hardware.&lt;/strong&gt; MDM tools (Jamf,
Intune) can flip a device to lost mode the moment it isn&#39;t returned
on schedule -- but only if offboarding actually triggers that,
instead of a spreadsheet update nobody acts on.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Buyer-side verification before cash changes hands.&lt;/strong&gt; Checking a
serial against a loss registry and confirming a device isn&#39;t
locked takes under two minutes and kills the resale value of stolen
inventory at the point of sale. Cheapest intervention in the whole
chain.&lt;/li&gt;
&lt;/ul&gt;
</content>
        
    </entry>
</feed>
