Analyzing Firmware Images logo

Analyzing Firmware Images

Community
trilwu
analyzing-firmware-images

Extract, analyze, and assess firmware images from embedded devices, IoT hardware, routers, and similar targets — filesystem extraction, hardcoded credential discovery, binary analysis across architectures, web interface review, network service enumeration, emulation, and cryptographic assessment. Use when analyzing a firmware update file, reviewing IoT device security, hunting for hardcoded secrets in device firmware, or assessing the attack surface of an embedded system.

Overview

Publishertrilwu
Repositorysecskills
Skill nameanalyzing-firmware-images
Stars
144
Forks
15
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 trilwu on GitHub. Read the source before you install it.

Installation

Install the Analyzing Firmware Images 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/trilwu/secskills.git /tmp/secskills
mkdir -p .claude/skills
cp -r /tmp/secskills/secskills-core/skills/analyzing-firmware-images .claude/skills/analyzing-firmware-images
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Analyzing Firmware Images 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 Analyzing Firmware Images 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 Analyzing Firmware Images 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.

Analyzing Firmware Images

Firmware is a frozen Linux (or RTOS) image, and its security froze with it. Hardcoded credentials, command injection in CGI scripts, unsigned update packages, and debug interfaces left enabled are not edge cases -- they are the baseline. The work is extraction, orientation, and then asking the same questions you would ask of any system, with the knowledge that nobody has patched this one since it shipped.

When to Use

  • Extracting and analyzing a firmware update file (.bin, .img, .chk, .trx)
  • Reviewing IoT device security posture from a firmware image
  • Looking for hardcoded credentials, keys, or secrets in device firmware
  • Assessing the attack surface of an embedded device or router
  • Evaluating update mechanisms and signature verification
  • Analyzing a bare-metal or RTOS image from a microcontroller

When NOT to Use

  • Standard x86/x64 binary reverse engineering -- use analyzing-binaries
  • Malware sample analysis -- use analyzing-malware
  • Source code is available -- use auditing-code-for-vulnerabilities

Firmware Acquisition

Before analysis comes acquisition. The method determines what you get.

MethodWhat you getNotes
Vendor downloadUpdate package, often compressed or encryptedCheck support portals, FTP servers, and FCC filings
OTA sniffingUpdate payload in transitmitmproxy or tcpdump on the device's update channel; many devices use plain HTTP
UART/serial consoleShell access, bootloader interactionThree wires (TX, RX, GND); identify with a multimeter or logic analyzer
JTAG/SWDFull memory read, debug accessRequires pin identification; JTAGulator, OpenOCD
Chip-offRaw flash contents (NAND/NOR)Desolder the flash chip; read with a programmer (CH341A, FlashcatUSB); last resort
Bootloader extractionDump via U-Boot md or sf read commandsIf the bootloader shell is accessible over UART

For OTA interception, configure the device to proxy through mitmproxy. Many devices ignore proxy settings -- ARP spoofing or a transparent bridge may be required. If the update is over HTTPS, check whether the device validates certificates at all; a surprising number do not.

Initial Analysis

Start with format identification and entropy analysis before extracting.

bash
file firmware.bin
binwalk firmware.bin              # identify embedded filesystems and compression
binwalk -E firmware.bin           # entropy analysis -- high entropy = compressed or encrypted
hexdump -C firmware.bin | head -64   # header bytes reveal container format
strings -n 10 firmware.bin | head -100  # quick orientation

What the entropy plot tells you:

  • Flat high entropy (close to 1.0) across the entire image -- encrypted or compressed as a unit. You need the decryption key or decompression method before you can proceed.
  • Regions of high entropy separated by low-entropy headers -- compressed filesystem partitions with metadata between them. Normal; extract the partitions.
  • Low entropy throughout -- uncompressed filesystem or raw flash. Direct extraction should work.

Common container formats:

Header magicFormat
hsqs / sqshSquashFS
UBI#UBI/UBIFS
0x1985JFFS2
0x28cd3d45cramfs
HDR0TRX (Broadcom routers)
\x27\x05\x19\x56uImage (U-Boot)

Extraction

Use the right tool for the filesystem. Generic extraction misses metadata and permissions.

bash
# General recursive extraction -- good starting point
binwalk -Me firmware.bin

# SquashFS -- the most common embedded filesystem
unsquashfs -d rootfs squashfs-root.img
# Non-standard SquashFS (vendor-modified): use sasquatch
sasquatch -d rootfs squashfs-root.img

# JFFS2
jefferson firmware.jffs2 -d rootfs

# UBI/UBIFS
ubireader_extract_images firmware.ubi
ubireader_extract_files firmware.ubi

# cramfs
cramfsck -x rootfs cramfs.img

# Raw NAND dumps may need OOB data stripped first
nandextract firmware.nand

If binwalk finds nothing and entropy is high, the image is likely encrypted. Look for a bootloader or earlier firmware version that contains the decryption routine. Some vendors ship the decryption key in the bootloader or in a companion partition.

Filesystem Analysis

Once extracted, treat the rootfs as a Linux system you are auditing for the first time.

Credentials and Secrets

bash
# Password files
cat rootfs/etc/passwd
cat rootfs/etc/shadow
# Hardcoded credentials -- these are endemic
rg -rn 'password|passwd|admin|root|default' rootfs/etc/ --include='*.conf'
rg -rn 'BEGIN (RSA|EC|OPENSSH|DSA) PRIVATE KEY' rootfs/
rg -rn 'api[_-]?key|secret[_-]?key|token' -i rootfs/

# Certificates and keys
find rootfs -name '*.pem' -o -name '*.key' -o -name '*.crt' -o -name '*.p12'
# WiFi and VPN credentials
find rootfs -name 'wpa_supplicant*' -o -name '*.ovpn' -o -name 'ipsec.*'

Default credentials in /etc/shadow are the single most common firmware finding. Check whether root has a password hash and whether it is crackable -- it usually is. See cracking-passwords for hash handling.

Configuration and Services

bash
# Startup scripts reveal what runs and how
ls rootfs/etc/init.d/ rootfs/etc/rc.d/
cat rootfs/etc/inittab
# Systemd units if present
find rootfs -name '*.service' -path '*/systemd/*'

# Installed packages and versions
cat rootfs/etc/opkg/status 2>/dev/null    # OpenWrt-based
ls rootfs/usr/lib/ipkg/info/ 2>/dev/null  # older ipkg

# Network configuration
cat rootfs/etc/network/interfaces 2>/dev/null
rg -rn 'iptables|ip6tables|nftables' rootfs/etc/

Web Interface

The web interface is where the exploitable bugs live. Embedded web servers are typically BusyBox httpd, lighttpd, uhttpd, or GoAhead, serving CGI scripts written in shell, Lua, or C.

bash
# Find the web root
find rootfs -name 'httpd*' -o -name 'lighttpd*' -o -name 'uhttpd*'
ls rootfs/www/ rootfs/usr/www/ rootfs/usr/share/www/ 2>/dev/null

# CGI scripts -- these are the attack surface
find rootfs -name '*.cgi' -o -name '*.sh' -path '*/cgi-bin/*'
find rootfs -name '*.lua' -path '*/luci/*' -o -name '*.lua' -path '*/www/*'

# Command injection patterns in CGI
rg -rn 'system\(|popen\(|exec\(|os\.execute|io\.popen|\`.*\$' rootfs/www/
rg -rn '\$QUERY_STRING|\$REQUEST_URI|\$HTTP_' rootfs/www/

Command injection through CGI parameters is endemic in embedded web interfaces. The pattern is a CGI script that takes user input from a query parameter and passes it to a shell command without sanitization. Review every CGI script for this pattern -- system(), popen(), backtick execution, os.execute(), and io.popen() with any user-controlled input.

Also check for:

  • Authentication bypass (pages accessible without login)
  • Cross-site scripting in diagnostic pages (ping, traceroute, DNS lookup)
  • Path traversal in file-serving handlers
  • Hardcoded session tokens or predictable session generation

Binary Analysis

Firmware binaries target non-x86 architectures. Identify the architecture before disassembly.

bash
# Identify architecture from ELF headers
file rootfs/usr/sbin/*
readelf -h rootfs/usr/sbin/httpd    # Machine field: ARM, MIPS, PowerPC

# Common architectures in firmware
# ARM (little-endian)  -- modern IoT, cameras, some routers
# MIPS (big-endian)    -- routers (Broadcom, Atheros, MediaTek)
# MIPS (little-endian) -- some Realtek-based devices
# PowerPC              -- older enterprise networking gear

Cross-Architecture Disassembly

Ghidra handles all common firmware architectures natively. Load the binary, select the correct processor and endianness, and auto-analyze.

bash
# Ghidra headless analysis
analyzeHeadless /proj FirmwareProj -import rootfs/usr/sbin/httpd \
  -processor ARM:LE:32:v7 -postScript DecompileAll.java

# radare2 with architecture specification
r2 -a arm -b 32 rootfs/usr/sbin/httpd
# For MIPS big-endian:
r2 -a mips -b 32 -e cfg.bigendian=true rootfs/usr/bin/target

Focus disassembly on:

  • The web server and CGI handler binaries
  • Custom daemons (anything not from BusyBox or standard packages)
  • Shared libraries that implement device-specific functionality
  • Binaries that run as root and accept network input

Emulation with QEMU

bash
# User-mode emulation for individual binaries
qemu-arm -L rootfs/ rootfs/usr/sbin/httpd
qemu-mipsel -L rootfs/ rootfs/usr/bin/target
# Use -strace to trace syscalls
qemu-arm -strace -L rootfs/ rootfs/usr/sbin/httpd

# If the binary needs specific /dev nodes or /proc, use chroot
sudo chroot rootfs/ /usr/sbin/httpd
# Or mount necessary filesystems
sudo mount -t proc proc rootfs/proc
sudo mount -t sysfs sysfs rootfs/sys

Network Services

Enumerate what the device exposes on the network by reading init scripts and binary configurations rather than by scanning a live device.

bash
# Services started at boot
rg -rn 'start\(\)|start_service' rootfs/etc/init.d/
# Listening ports from configuration
rg -rn 'listen|bind|port' rootfs/etc/ --include='*.conf'

# Common embedded services to look for
find rootfs -name 'telnetd' -o -name 'dropbear' -o -name 'sshd'
find rootfs -name 'upnpd' -o -name 'miniupnpd' -o -name 'minissdpd'
find rootfs -name 'mosquitto*' -o -name 'mqtt*'
find rootfs -name 'snmpd' -o -name 'snmp.conf'

Common findings:

  • Telnet enabled with default or no credentials
  • UPnP/SSDP exposing internal service descriptions to the WAN
  • MQTT brokers with no authentication
  • SNMP with default community strings (public, private)
  • Custom management protocols on non-standard ports with no authentication
  • TR-069 (CWMP) interfaces exposed beyond the ISP management VLAN
  • Debug ports (GDB server, serial-over-network) left active in production

Full-System Emulation

When individual binary emulation is insufficient, emulate the entire firmware.

bash
# FirmAE -- automated full-system emulation
sudo python3 firmae.py -r <brand> firmware.bin

# EMBA -- comprehensive firmware analysis framework
sudo ./emba -f firmware.bin -l ./logs

# firmwalker -- static analysis without emulation
./firmwalker.sh rootfs/

FirmAE and EMBA handle the hard parts: inferring the correct QEMU machine type, setting up the network, and patching /dev/ nodes. If they fail, manual QEMU system emulation requires building the correct device tree and kernel for the target platform.

Once a full system is running, test it as you would any networked service: scan with nmap, test the web interface, fuzz the custom protocols. The difference is that you have the filesystem and can read the code while you test.

Cryptographic Analysis

Firmware images routinely contain cryptographic material and implement custom cryptographic schemes.

bash
# Find encryption keys and certificates
find rootfs -name '*.pem' -o -name '*.der' -o -name '*.key' -o -name '*.pub'
rg -rn 'AES|DES|RSA|SHA256|MD5|encrypt|decrypt' rootfs/usr/lib/ --include='*.so'

# Check update signature verification
# Look for signature checks in the update handler
rg -rn 'verify|signature|sign|openssl|gpg' rootfs/usr/sbin/ rootfs/etc/init.d/

What to look for:

  • Unsigned firmware updates -- if the update mechanism does not verify a cryptographic signature, an attacker with network position can push arbitrary firmware. This is critical.
  • Symmetric-only update signing -- a shared key embedded in the firmware itself means anyone with the firmware can sign updates. Extract the key and demonstrate.
  • Hardcoded TLS certificates and private keys -- every device ships with the same key pair. Extract and demonstrate that one device's key decrypts another device's traffic.
  • Weak or custom encryption -- XOR "encryption" of configuration files, hardcoded keys for "encrypting" passwords, custom obfuscation routines that are not encryption at all.
  • Entropy sources -- embedded devices often have poor entropy at boot. Check whether /dev/urandom is seeded properly, whether the RNG is initialized before key generation, and whether the device has a hardware RNG that is actually used.

See reviewing-cryptography for detailed cryptographic review methodology.

Defensive Review Checklist

After completing the analysis, evaluate against these controls. The absence of any item is a finding.

ControlCheckCommon failure
Signed updatesIs the update package cryptographically signed with an asymmetric key? Is the signature verified before flashing?No signature at all, or symmetric-only
Secure bootDoes the bootloader verify the kernel and rootfs integrity?U-Boot with no signature verification
No default credentialsDoes the device force a password change on first use?root:root, admin:admin, or blank passwords
Minimal servicesAre only necessary services enabled?Telnet, UPnP, SNMP enabled by default
Debug interfacesAre UART, JTAG, and SSH disabled or locked in production?UART shell with root access, no authentication
Encrypted storageAre credentials and keys stored encrypted at rest?Plaintext passwords in config files
TLS everywhereDo all network services use TLS with valid certificates?Plain HTTP for web management, plain MQTT
Input validationDo CGI and API handlers validate and sanitize input?Direct shell injection through web parameters
Least privilegeDo services run as non-root where possible?Everything runs as root
Logging and auditAre security events logged?No logging, or logs only in volatile memory

Rationalizations to Reject

  • "It's on an isolated network." Embedded devices get exposed -- through UPnP, through misconfigured firewalls, through the cloud management portal that phones home. Assume reachability.
  • "The firmware is encrypted, so we can't analyze it." The decryption key is in the bootloader or in a previous unencrypted version. Encryption without secure boot is obfuscation, not protection.
  • "It's just a consumer device." Consumer devices form botnets. The same vulnerability classes apply.
  • "We can't patch it, so why report it?" Document it so the risk is understood and compensating controls can be applied.
  • "The vendor says it's secure." Vendor attestation is not evidence. The filesystem is evidence.
  • "It's a custom RTOS, not Linux, so standard tools won't work." The architecture-level tools (Ghidra, QEMU, binwalk entropy) work on any binary. Adapt the methodology, do not skip the analysis.
  • "Nobody would bother attacking this device." Mirai scanned the entire IPv4 space for default credentials. The bar for "bother" is a single script.

References

  • analyzing-binaries -- disassembly and reverse engineering methodology
  • auditing-code-for-vulnerabilities -- source-level review when code is available
  • enumerating-network-services -- service discovery and assessment
  • reviewing-cryptography -- cryptographic implementation review

Frequently asked questions

What does the Analyzing Firmware Images AI skill do?

Extract, analyze, and assess firmware images from embedded devices, IoT hardware, routers, and similar targets — filesystem extraction, hardcoded credential discovery, binary analysis across architectures, web interface review, network service enumeration, emulation, and cryptographic assessment. Use when analyzing a firmware update file, reviewing IoT device security, hunting for hardcoded secrets in device firmware, or assessing the attack surface of an embedded system.

Why use Analyzing Firmware Images on TypingMind?

Because you install it once and use it with any model. Analyzing Firmware Images 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 Analyzing Firmware Images in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/trilwu/secskills/tree/main/secskills-core/skills/analyzing-firmware-images. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Analyzing Firmware Images?

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 Analyzing Firmware Images?

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

Is the Analyzing Firmware Images AI skill free?

Yes. It is published on GitHub by trilwu 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 👇