Ablation is a self-contained reverse engineering framework.
It depends on neither legacy tools nor an MCP server, which removes the attack surface associated with MCP.
When used with Codex or Claude Code, it operates as an autonomous reverse engineering system.
Claude Code Setup & Configuration
Quick Start
1. Launch Session
cd ~/ablation && claude
2. Set Model
/model claude-sonnet-4-6
or
/model claude-sonnet-4-5
3. Define Objective
/goal reverse engineer [whatever you want]
4. Check Effort Level
Not sure what effort level to be set on? Ask inline without interrupting the main context:
/btw What effort level should we be set on right now?
Recommended Configuration (/config)
Auto-compact true
Continue automatically at usage limit true
Switch models when a message is flagged Switch automatically
Show tips false
Claude-drafted feedback off
Reduce motion false
Thinking mode true
Fast mode (Opus 5.5) false
Prompt suggestions true
Session recap true
Rewind code (checkpoints) true
Dynamic workflows true
Ultracode keyword trigger false
Dynamic workflow size unrestricted
Artifacts false
Verbose output true
Terminal progress bar true
Show turn duration true
Time format auto
Default permission mode Manual
Worktree base ref fresh
Use auto mode during plan true
Respect .gitignore in file picker true
Skip the /copy picker false
Copy on select true
Auto-scroll true
Open agents view by default false
← opens agents false
Auto-update channel disabled (config)
Theme Dark mode (colorblind-friendly)
Local notifications Auto
Push when actions required true
Push when Claude decides true
Output style Explanatory
Language Default (English)
Editor mode normal
Question auto-continue timeout never
Show last response in external editor true
Show PR status footer true
Model claude-sonnet-4-6
Auto-connect to IDE (external terminal) true
Claude in Chrome enabled by default true
Enable Remote Control for all sessions false
Dialog expiry never
Messages from your other sessions accept
Capabilities
Core Analysis Engine
| Capability | |
|---|---|
| On-Demand Analysis | Loads in seconds because it only analyzes the code being examined. A 2 GB firmware image opens as fast as a small one because nothing is parsed until it is needed. |
| Binary Lifting | Translates machine code to readable pseudo-C across 31 architectures. ARM64 and x86-64 get register-level taint tracking so data flow is visible in the lifted output. |
| Semantic Code Search | Finds functions by describing what they do because stripped binaries have no symbols to search. Queries are scoped to the target binary so results return in under a second. |
| Hypothesis Engine | Tracks competing theories about what a function does and scores evidence against each one. Sessions persist across engagements so analysis resumes where it left off. |
| DAG Adapter Language | Every architecture shares the same analysis code path because each ISA compiles to a directed acyclic graph. Adding a new ISA takes one grammar file, so the architecture count grows without multiplying the codebase. |
| Entropy Mapper | Maps binary sections by Shannon entropy to classify encrypted, compressed, and plaintext regions before analysis begins. Sections approaching H=8.0 indicate AES or compressed data so analysis effort concentrates on decrypted plaintext regions first. |
Learning & Continuity
| Capability | |
|---|---|
| The Flywheel | Every confirmed finding replays as a search pattern on every future binary. A vulnerability confirmed on one vendor's firmware surfaces automatically when sweeping a different vendor's binary because confirmed findings are stored and replayed at engagement start. |
| Cross-Target Learning | Extracts a transferable signature from every confirmed finding and stores it in the pattern library so the tool grows sharper across targets. The Flywheel replays those signatures; this module generates them. |
| KASAN-Oracle | Scans a KASAN debug build first because debug builds expose control flow that stripping removes. Confirmed candidates from the debug build then validate against the stripped production binary, so the debug image acts as a ground-truth oracle rather than a secondary target. |
Vulnerability Discovery & Validation
| Capability | |
|---|---|
| Taint Analysis | Follows network input through the call graph to security-sensitive sinks across 16 architectures. Library call sites do not stop the trace because register-level tracking resolves shared library stubs. |
| Cross-Library Analysis | Traces an argument across up to three shared library hops so taint paths that cross library boundaries reach their sink rather than stopping at the first call. |
| Pre-Authentication Exposure | Finds every code path from the network entry point to a security-sensitive operation that runs before authentication. Manual review starts at the authentication boundary because the output is a reachability map, not a raw disassembly. |
| Sink Arg Classifier | Classifies what value reaches each exec-class sink so a taint hit is either confirmed as externally controlled or eliminated as a constant before manual review begins. |
| Finding Validation | Generates a working HTTP request that reproduces each confirmed finding at the protocol level. The request includes the expected response delta and CWE classification so a PSIRT submission arrives with working proof. |
| Compiler Defect Detection | Finds security defects that the compiler introduced through miscompilation. Optimizer defects that silently select the wrong comparison operand can produce hundreds of confirmed vulnerabilities across a single firmware image. |
| Source Code Auditing | Scores every source file by risk and batches the low-risk ones so manual review covers only the files most likely to contain vulnerabilities. |
| Cryptographic Analysis | Recovers cryptographic key material directly from compiled firmware and audits TLS and JWT configurations, so cryptographic findings do not require source access or a decryption oracle. |
| Protocol FSM Extractor | Extracts the handshake state machine from network protocols in stripped ARM64 binaries using static loop and BSS state-variable analysis. Authentication state transitions become explicit without a running server or traffic capture. |
| BMP LSB Key Extractor | Recovers cryptographic key material hidden in BMP pixel LSBs using a Shamir-style Lagrange polynomial secret-sharing scheme. Firmware and APK packages that conceal keys in image assets are handled without prior knowledge of the encoding. |
| Format String Scanner | Identifies format string vulnerabilities across the full printf family, syslog, and custom logging functions by tracing format argument provenance in x86-64 ELF binaries. Every call site where the format argument is not a .rodata string literal is a finding. |
| Heap Vulnerability Scanner | Detects heap use-after-free, double-free, and overflow patterns in x86-64 and LoongArch64 binaries using a forward symbolic pass tracking pointer state across allocation and free operations. Static analysis produces heap findings in firmware that cannot be fuzz-tested. |
| SQL Injection Scanner | Detects raw SQL injection in C/C++ binaries that call the MySQL C API or SQLite3 directly without a prepared-statement layer. Chains format-string builder analysis to SQL sink tracing so snprintf-to-mysql_query patterns surface that a format string scanner alone would miss. |
| RPC Server Analyzer | Enumerates Windows RPC server interfaces and flags those registered with RPC_C_AUTHN_LEVEL_NONE so any caller can reach the interface without credentials. Pairs with Windows PE Security Analysis for RPC coverage. |
| IOCTL Attack Surface | Maps the IOCTL dispatch surface of a Windows kernel driver and flags METHOD_NEITHER codes before manual review begins. An integer-overflow path from a user-supplied length to a kernel allocation is a HIGH finding. |
| Length Underflow Scanner | Detects integer underflow in protocol parsers where a user-supplied length field has a fixed header size subtracted, wraps on truncation, and reaches a downstream allocation or copy without a prior bounds check. |
| Setuid Auth Oracle | Identifies setuid binaries that call getspnam and crypt without PAM mediation, exposing a brute-force oracle because exit-code-based result disclosure bypasses rate limiting that pam_tally2 and faillock enforce. Absence of pam_start in the PLT surfaces candidates without a taint run. |
| Lua Injection Analyzer | Detects unsanitized user-controlled strings passed to a Lua runtime via Python format operations. Built from two CRITICAL findings in FortiAnalyzer and FortiManager where no other scanner detected the class. |
Specialized Recovery
| Capability | |
|---|---|
| C++ Vtable Reconstruction | Recovers C++ virtual dispatch tables from stripped ELF binaries on any architecture and names each slot. A generated type-annotation script propagates slot names through every call site so the decompiler shows method names instead of function pointer offsets. |
| PPC32 GOT2 Resolver | Resolves indirect calls in stripped GOT2-PIC PPC32 binaries at 98.6% coverage so taint paths through BCTRL stubs do not dead-end at the first unresolved call. |
| Dynamic Sandbox | Executes leaf functions in an emulated sandbox to fingerprint behavior where static analysis alone cannot distinguish branches. |
| Firmware Extraction | Decrypts vendor-encrypted firmware without a key because predictable patterns in inner partitions reveal it. Handles vendor-modified squashfs images including ARM64 BCJ-filtered variants. |
| Firmware Key Corpus | Classifies recovered firmware keys against a cross-image family table. An anomalous key count flags a potential finding before decryption begins. |
Platform & Runtime Support
| Capability | |
|---|---|
| Windows PE Security Analysis | Audits a Windows PE binary for security weaknesses in a single pass. It checks CFG bypass exports, SafeSEH gaps, unauthenticated RPC endpoints, COM registration gaps, and kernel pool overflows so the attack surface is visible before manual review begins. |
| Windows Kernel Driver & BYOVD | Classifies signed kernel drivers for BYOVD capability before they are deployed. Physical memory access, privilege token manipulation, and callback removal each trigger a finding. |
| Android / APK Analysis | Maps the APK attack surface and ranks native libraries by risk before manual analysis begins. |
| HarmonyOS / ArkTS Analysis | Analyzes HarmonyOS HAP packages and ArkTS bytecode so OpenHarmony targets can be assessed without source access. Surfaces exported abilities, permission grants, and native library attack surface before manual review begins. |
| Go Binary Reverse Engineering | Recovers function names from stripped Go binaries because the runtime embeds a metadata table that survives stripping. |
| Erlang / BEAM Analysis | Decodes BEAM bytecode to readable pseudo-IR so Erlang and Elixir applications can be analyzed without source. High-risk import patterns and obfuscation indicators surface without executing the application. |
| MicroPython .mpy Lifter | Lifts .mpy v6 bytecode to readable pseudo-Python so MicroPython firmware on ESP32, STM32, and CC13xx targets can be analyzed without source. Flags dangerous imports and machine-level memory access calls in the lifted output. |
| Game & Legacy RE | Labels stripped game binary functions by engine so analysis starts at the game logic layer instead of the engine layer. Also handles PS3 Cell SPU and classic Mac OS PEF binaries. |
| PS3 SELF Decryptor | Decrypts PS3 SELF (Signed ELF) files to plain ELF because lief and BinaryContext reject SCE-format magic. Keyset detection is automatic from the key_rev field so no manual key lookup is needed. |
| MCP Server Security Scanner | Scans MCP server source for unauthenticated endpoints, exec-class methods in allowed tool lists, and network-wide socket bindings. All three issue classes appeared in production MCP servers where no other scanner detected them. |
| LoongArch64 | The only public reverse engineering toolkit built specifically for LoongArch64. Every analysis phase works without architecture-specific workarounds. |
| nanoMIPS | Follows tainted data through nanoMIPS binaries interprocedurally, resolving branch targets and PLT stubs that no other public tool handles. |
Automotive ECU Analysis
| Capability | |
|---|---|
| ECU Architecture Detector | Identifies the ISA of a flat ECU ROM from raw bytes alone. PPC32, SH-2A, TriCore, RH850, V850, M68K, C28x, and M32R all produce recognizable structural signatures without symbols or file headers. |
| ROM Layout Analyzer | Splits a flat ROM into CODE, CALIBRATION, and IVT regions before analysis begins. The interrupt vector table is excluded from function discovery so taint trackers start from clean application code boundaries. |
| Calibration Parsers | Reads TunerPro XDF, ASAM A2L, RomRaider XML, open_damos, ConesCAN, and Nissan Pytrex definition files and injects human-readable table labels into BinaryContext. Every array in the ROM has a name when the taint tracker reaches it. |
| CAN DBC Parser | Reads CAN signal database files and makes message and signal definitions available during binary analysis. No other RE tool brings CAN signal context into the taint analysis pipeline. |
| ECU Security Access Scanner | Audits UDS SecurityAccess implementations for weak seed generation, static keys, and bypassable challenge-response logic. Runs before taint analysis so weak authentication findings are confirmed before deeper code tracing begins. |
| TriCore Taint Analysis | Follows attacker-controlled data to security-sensitive sinks in Infineon TriCore AURIX TC1.6x ECU firmware using the GHS EABI and IAR EABI calling conventions. Automotive targets on TriCore receive the same interprocedural taint coverage as ARM64 and x86-64. |
| RH850 Taint Analysis | Follows attacker-controlled data through Renesas RH850 ECU firmware using the GHS CC-RH and IAR RH850 ABIs. Flat ROM support handles images shipped without ELF headers because most automotive RH850 firmware ships without them. |
Workflow & Quality Gates
| Capability | |
|---|---|
| Version Diffing | Checks whether the logic changed, not whether the file changed, so a cosmetic recompile cannot hide an unpatched vulnerability. |
| Kernel Module Subsystem Classifier | Classifies a Linux kernel module by its source tree path to determine attack surface before analysis begins. Network-reachable modules surface first so manual review starts at the highest-risk code. |
| Module Quality Gate | FORGE audits every user-written module before it can be stored, so local extensions meet the same standard as the modules that ship with Ablation. Every HIGH and CRITICAL finding is tagged with the pipeline stage where it would surface without the audit, from a free static-check catch at develop time to an 18x-cost find in production. |
| Coverage Gate | The taint engine requires 85% statement coverage on every commit, so a regression in taint logic fails the test suite before it reaches a firmware engagement. |
| YARA Rule Generation | Synthesizes YARA detection rules from PE and ELF scanner findings so confirmed vulnerability signatures are immediately deployable in detection pipelines without manual rule authoring. |
31 Decompilers
Native CPUs
| ISA | Variants |
|---|---|
| x86 | x86-32 · x86-64 |
| ARM | ARM-32 · ARM-64 · Thumb |
| MIPS | MIPS-32 · nanoMIPS · MIPS-64 |
| PowerPC | PPC-32 · PPC-64 · VLE |
| RISC-V | RV32 · RV64 · HiSilicon ext |
| ARC | ARC EM/HS |
| LoongArch | LoongArch64 |
| Intel MCS | 8051 |
Automotive ECUs
| ISA | Variants |
|---|---|
| V850 / RH850 | V850-32 · RH850 |
| Renesas SH | SH-2A |
| Motorola | M68K / CPU32 · MC88100 |
| Infineon | TriCore TC1.6x |
| TI DSP | TMS320C28x |
VMs & Runtimes
| ISA | Variants |
|---|---|
| DEX | Dalvik · ART |
| ARK | ArkTS |
| BEAM | Erlang · Elixir |
| MicroPython | .mpy v6 |
Project Background
Real-World Results
Real-World Results
Ablation has been used to analyze production firmware and kernel drivers from Cisco, Fortinet, TencentOS, Huawei, and more.
Following coordinated disclosure, the Cisco Product Security Incident Response Team (PSIRT) adopted Ablation for internal vulnerability triage and continues to use it to evaluate ongoing disclosure reports. Cisco Adaptive Security Appliance (ASA) LINA is among the artifacts analyzed, with findings submitted through CERT/CC VINCE.
| CVE | Product | Title | CVSS |
|---|---|---|---|
| CVE‑2026‑76420 | Secure Firewall Management Center (FMC) | Peer Impersonation | 9.0 Critical |
| CVE‑2026‑76412 | Secure Firewall Management Center (FMC) | Privilege Escalation to root | 8.5 High |
| CVE‑2026‑76413 | Secure Firewall Management Center (FMC) | Single Sign-On Token Forgery | 8.5 High |
| CVE‑2026‑76447 | Identity Services Engine (ISE) | OCSP Responder Authentication Bypass | 5.3 Medium |
Acknowledgments
Acknowledgments
The code couldn't exist without their ideas. Though no code was written by these authors, doctors, and engineers their work has direct influence on the production of code that makes Ablation what it is.
Research Papers
Books supplied by www.oreilly.com | github.com/oreillymedia
| Title | Authors |
|---|---|
| The Art of Software Security Assessment | Mark Dowd, John McDonald, Justin Schuh |
| Practical Binary Analysis | Dennis Andriesse |
| Practical Malware Analysis | Michael Sikorski, Andrew Honig |
| Practical Reverse Engineering | Bruce Dang, Alexandre Gazet, Elias Bachaalany |
| Hacking: The Art of Exploitation (2e) | Jon Erickson |
| Learning Linux Binary Analysis | Ryan O'Neill |
| Windows Internals Part 1 & 2 | Pavel Yosifovich, Mark Russinovich, David Solomon, Alex Ionescu, Andrea Allievi |
| Rootkits: Subverting the Windows Kernel | Greg Hoglund, Jamie Butler |
| Advanced Compiler Design and Implementation | Steven Muchnick |
| Engineering a Compiler | Keith Cooper, Linda Torczon |
| Practical IoT Hacking | Fotios Chantzis, Ioannis Stais, Paulino Calderon, Evangelos Deirmentzoglou, Beau Woods |
| Inside the Android OS: Building, Customizing, Managing and Operating Android System Services | G. Blake Meike |
| Malware Analysis and Detection Engineering | Abhijit Mohanta, Anoop Saldanha |
| Evasive Malware | Kyle Cucci |
| Hacking Cryptography | Kamran Khan, Bill Cox |
| Real-World Cryptography | David Wong |
Honorable Mention
Microsoft Excel (Data Analysis ToolPak) When analyzing closed infrastructure or securing black-box systems, this exact process is called timing analysis or telemetry reverse engineering. Without source code, the Data Analysis ToolPak mathematically deconstructs how an application works on the backend by strictly observing its inputs and outputs.
Getting Started
Install
Install
Do not run
pip install ablation. An unrelated package occupies that name on PyPI. You will install the wrong thing and see no error.
From PyPI:
pip install ablation-tool
With LLM features (anthropic client):
pip install "ablation-tool[llm]"
Verified install (pinned commit SHA) — pip rejects any substitution:
pip install -r requirements-verified.txt
Local / development:
git clone https://github.com/Ablation-Tool/ablation.git ~/ablation
pip install -e ~/ablation/
Verify the install:
python -c "import ablation; print(ablation.__version__)"
python -c "from ablation.analyzers.binary_context import BinaryContext; print('ok')"
Responsible Use
Responsible Use
Ablation is built for authorized security research. Use it only against systems you own or have explicit written permission to test. Running it against systems without authorization violates computer fraud laws in most jurisdictions. The author is not responsible for misuse.
Standards
Standards
This defines how software should be built, organized around four practice areas: preparing the development environment, protecting the codebase from unauthorized changes, writing software with security built into every layer, and responding when vulnerabilities surface after release.
What this means for Ablation in practice is a mandatory ten-section security audit on every module before it touches the repository.
The audit covers input validation, memory handling, integer safety, error handling, concurrency, dependency security, injection prevention, secrets handling, logging, and residual risks.
No module is committed without passing it.
Taint trackers, which are the modules that matter most if they get something wrong, carry an additional requirement: 85% statement coverage enforced by a hard gate in the test suite, verified against byte-exact synthetic instruction sequences that test ABI register behavior, not just whether the code runs without crashing.
Dependencies are pinned with exact hashes and scanned weekly for known CVEs.
CISA asks software manufacturers to commit to three things: take ownership of the security outcomes of the people who rely on the software, be transparent about its limitations, and treat security as a requirement rather than a recommendation.
For a RE framework, taking ownership of security outcomes has a specific meaning.
A taint tracker that misses a vulnerability does not just produce a wrong answer.
It produces a result someone may act on as evidence the binary is clean.
That is why a false negative in any Ablation taint tracker is classified as a CRITICAL defect.
Every tracker ships with a documented known-anomalies file describing the exact conditions under which it can produce false negatives, so a researcher knows when to apply manual verification on top of the automated result.
The precommit gate is not advisory. No code enters the repository without passing it.
The US Department of Defense handbook for reverse engineering methodology.
It defines four phases:
Phase 1. Every engagement starts with a session file recording the target description, artifact hash, tool selection, and documentation plan before any analysis runs. The rule that semantic search runs before manual disassembly maps to the following phase.
Phase 2 Requires characterizing a binary at the function level before doing detailed trace work. The DynamicSandbox covers the following.
Phase 3. The structured retrospective written at the close of every completed project maps to the final phase.
Phase 4. The standards document tracks 29 specific requirements drawn from all three frameworks. All 29 are met.
1 maintainer
Metadata
1 maintainer
Release files for ablation-tool 2.81.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| ablation_tool-2.81.0.tar.gz | 2.3 MB | With provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| ablation_tool-2.81.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 4.8 MB
Release files / ablation_tool-2.81.0.tar.gz
| Download URL | ablation_tool-2.81.0.tar.gz |
|---|---|
| Size | 2.3 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
69bfd825fbfb259cc43b97d5f589f26f3fba2ba11a36a7bd5ed2a687c048e461
|
|
BLAKE2b-256 checksum How to use checksums |
40250130451eecf18b6e461ff1fd83acbdf24a950b9957463e2835c4011b5a12
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 15, 2026.
Transparency logRelease files / ablation_tool-2.81.0-py3-none-any.whl
| Download URL | ablation_tool-2.81.0-py3-none-any.whl |
|---|---|
| Size | 2.4 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3d4c68c193558f1a2023f2d8e5998b43f7a353d9065e99b14e6aaec289bbb4f3
|
|
BLAKE2b-256 checksum How to use checksums |
5535025bb47fd85f548d70b29817355209ebf903d81075c44e03e5995140c4f4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 15, 2026.
Transparency logSecurity for ablation-tool 2.81.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for ablation_tool-2.81.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 15, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
