8000 GitHub - Ablation-Tool/ablation: Ablation is an AI-assisted reverse engineering (AIARE) framework · GitHub
[go: up one dir, main page]

Skip to content
Ablation-ToolPublic

About

Ablation is an AI-assisted reverse engineering (AIARE) framework

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

5 stars

Watchers

0 watching

Forks

Repository files navigation

ABLATION

NIST SP 800-218 CISA Secure by Design MIL-HDBK-115C

ABL∀TION is an AI-assisted reverse engineering (AIARE) framework.


GEM native workbench demo

ZOE · allgems

ZOE allgems orbital eye emblem

ZOE is the GEM, but the GEM is not ZOE. ZOE is allgems. The complete ZOE component is included in zoe/: its native Cognitive Field, Bayesian memory, geometry, replayable knowledge graph, unified hub, and the other GEM source trees. This repository supplies Ablation's core engine and native Lab as the host component; they are not duplicated inside zoe/components/. The imported ZOE revision is recorded in zoe/ZOE-SOURCE.json.

Install ZOE's native hub alongside Ablation and GEM:

python -m pip install ./zoe/guiai-lab
export ZOE_ROOT="$PWD/zoe"
zoe
zoe list
zoe launch ablation -- --help

zoe opens the unified hub without starting another engine or an AI agent. Component launches are explicit. The ablation and gem-lab entries resolve to this host checkout; the other upstream sources remain under zoe/components/. ZOE_ABLATION_ROOT can override host detection when ZOE is used from another layout. Source presence does not imply runtime verification or permission to act. See ZOE's architecture and verification guide and its historical recognition and attribution.

Capabilities

ABL∀TION's Default GEM
Capability Code Description
a11y Optimization 1 2 Auto-detects & merges your accessibility settings & tools into the GUI
Accessible Volume 1 2 Measures and records next-token forecast breadth with a local model
Agent Bridge 1 2 Embeds your AI Agent into the GEM
Debug Diagnostics 1 2 Shows dependency and terminal diagnostics before requesting troubleshooting permission
Evidence Integrity 1 2 3 Tracks reviews and validates evidence targets
GEM Verification 1 2 3 Verifies and replays spatial investigation evidence
Lab Navigation 1 2 3 Searches and visualizes discoveries while restoring sessions, agents, and preferences
OS Optimization 1 2 Recognizes the operating system in use and adapts GEM to that environment
OS Support 1 2 3 Supports Linux, Windows, macOS, adaptive presentation, and safe installation
Decompilers, Disassemblers & Analyzers

Native CPUs

ISA Variants
Alpha AXP Alpha 21064 · 21164 · 21264 · ECOFF · ELF
ARC ARC EM · ARC HS
ARM ARM-32 · ARM-64 · Thumb
Intel MCS 8051 / MCS-51 · CC254x
LoongArch LoongArch64 · KASAN/KCOV V2
MIPS MIPS-32 · nanoMIPS · MIPS-64
PowerPC PPC-32 · PPC-64 · VLE · PPC476 Book E · XCOFF32 / AIX
PA-RISC PA-RISC 1.0 · 1.1 · HP-UX SOM · DLT/PLT · Unwind
RISC-V RV32 · RV64 · HiSilicon ext
S/390 Big-endian S/390 · z/Arch subset · LE/C++ PPA · XPLINK · BFP · Pseudo-C
x86 x86-32 · x86-64
Xilinx MicroBlaze · Versal PDI/CDO · ZynqMP BOOT.BIN

Automotive ECUs

ISA Variants
Infineon TriCore TC1.6x
Infineon C166 C166 · C167
Motorola M68K / CPU32 · MC88100
Renesas SH SH-2A
TI DSP TMS320C28x
V850/RH850 V850-32 · RH850

VMs & Runtimes

ISA Variants
ARK ArkTS
BEAM Erlang BEAM bytecode
DEX Dalvik · ART · Java-like pseudocode · JNI
Java .class · JAR
MicroPython .mpy v6
Python .pyc (3.8–3.12)
Core Engine
Capability Code Description
ABI Models 1 2 Normalizes x86 instructions and models register aliases, calling conventions
Analysis Caching 1 2 Reuses analysis results and session state across runs
Architecture DSL 1 Describes ISA encodings and semantics as reusable graphs
Argument Tracking 1 2 Tracks registers and sink arguments across calls
ARM Disassembly 1 2 Provides targeted disassembly and dataflow analysis
ARM64 Slicing 1 2 Traces ARM64 store addresses backward
Binary Parsing 1 2 Parses ELF, PE, and Mach-O metadata and inspects ELF hardening
Cryptographic Analysis 1 2 Detects cipher implementations, recovers XOR and affine keys
Data-Flow Analysis 1 2 Builds control-flow graphs, propagates dataflow facts
Entropy Mapping 1 Locates compressed and encrypted regions using entropy
Expression Analysis 1 Converts instructions into typed expressions and merges computations
Function Discovery 1 2 Recovers function entries from stripped binaries and ROMs
Function Emulation 1 Emulates ARM64 and x86-64 leaf functions
Function Prioritization 1 2 Profiles calls and filters findings in functions
Intermediate Representation 1 Lifts machine code into shared intermediate representations
Knowledge Graph 1 2 Combines knowledge graphs with agent-driven retrieval
SSA Modeling 1 2 Models register definitions using static single assignment
Symbol Recovery 1 2 Recovers ELF function addresses and names from unwind metadata, DWARF
ReAct (Reasoning + Acting)
Capability Code Description
AI-Guided Analysis 1 2 Pre-fetches function context for LLM analysis
Approximate Matching 1 2 Indexes opcode sequences for approximate matching
Cross-Target Learning 1 2 Reuses findings and vendor sink patterns
Evidence Validation 1 2 Ranks probes by expected information gain and executes approved adapters
Function Identification 1 2 Names known functions and preserves analyst names
Hypothesis Tracking 1 Ranks and preserves vulnerability hypotheses
Pattern Mining 1 2 Mines CWE patterns and scans binaries
Semantic Search 1 Ranks functions using natural-language embeddings
Security Research
Capability Code Description
Attack-Surface Detection 1 Prioritizes network-reachable propagated-argument sinks
Bounds Checking 1 2 Detects underflows and validates chunk bounds
Call Classification 1 2 3 Classifies sink arguments and filters constant-format calls
Call Wrappers 1 2 Discovers sink wrappers and profiles sanitization coverage
Compiler Hardening 1 2 Detects miscompilations and inconsistent binary hardening
Driver Vulnerabilities 1 2 Ranks BYOVD primitives and identifies unsafe IOCTL handlers
Engine Analysis 1 Identifies game engines and RCE surfaces
Exploit Scoring 1 2 Builds shellcode and computes cyclic offsets
Extension Security 1 Detects insecure externally connectable browser extensions
Format-String Vulnerabilities 1 Finds non-constant format arguments with provenance
Fortinet Surfaces 1 2 Reconstructs Apache handlers and Fortinet execution sinks
Function Checks 1 2 Classifies risky calls and deprecated functions
Heap Vulnerabilities 1 2 Detects allocation overflows, use-after-free, double-free
Injection Analysis 1 2 Classifies SQL and Lua injection provenance
KASAN Validation 1 Verifies KASAN candidates in stripped binaries
Path Validation 1 2 Checks path feasibility and stores reviewed findings
Source Analysis 1 2 Profiles sources and traces dangerous sinks
Taint Analysis 1 2 3 Traces attacker-controlled data across architectures
Unauthenticated Exposure 1 Maps unauthenticated routes and reachable sinks
Vulnerability Sweeps 1 2 Runs resumable pre-filtering, sink classification, and taint analysis
Firmware Analysis
Capability Code Description
Architecture Detection 1 2 Identifies ROM architectures and classifies regions
bFLT Loader 1 2 3 Loads bFLT executables and applies ARM32 relocations
C166 Toolkit 1 2 Analyzes C166 control flow and CAN updates
CAEN Containers 1 Decodes CAEN firmware and verifies additive checksums
ECU Calibration 1 2 Parses ECU calibration and scaling metadata
ECU Security 1 2 Detects predictable seeds and traces diagnostic writes
Embedded Decoding 1 2 Decodes C166 and C28x instructions
Firmware Steganography 1 Extracts cryptographic material from firmware images
Firmware Triage 1 2 Scans firmware and extracts vendor partitions
Fortinet 1 2 Recovers FortiOS keys and firmware containers
HP-UX SOM 1 2 3 Recovers SOM dynamic linkage, unwind ranges, and PA-RISC calls
JFFS2 Recovery 1 Reconstructs little-endian JFFS2 files with CRC and extent validation
Juniper 1 2 3 4 5 6 7 Verifies Junos packages and analyzes MIPS functions
Legacy Formats 1 2 3 Parses M88K COFF, DG/UX ELF, and Guardian PAK archives
Texas Instruments 1 2 Parses TI COFF2 symbols and embedded models
TIMI Saves 1 2 3 4 Verifies QSYS SAVF records and extracts objects
WDB Reader 1 Streams Xilinx waveform events with bounded memory
Platform & Runtime Analysis
Capability Code Description
Android Cross-Layer 1 2 3 4 5 Analyzes Android packages, JNI bindings, and native sinks
C++ Calls 1 2 Reconstructs C++ hierarchies and virtual call graphs
Erlang/MicroPython 1 2 Lifts BEAM bytecode and flags dynamic operations
Go Binaries 1 2 Recovers Go names, handlers, and message framing
HarmonyOS Apps 1 2 3 Reconstructs ArkTS pseudocode and embedded endpoints
IBM AIX 1 2 Parses XCOFF sections and BFF archives
IBM z/OS 1 2 3 4 5 6 7 Decode z/OS binaries and compare builds
Java/Spring 1 2 Parses Java artifacts and detects deserialization gadgets
LoongArch/nanoMIPS 1 2 3 Decodes and analyzes LoongArch and nanoMIPS
PowerPC Calls 1 2 3 Resolves PowerPC indirect calls and trampolines
Runtime Security 1 2 3 Audits Python, Perl, and Electron code
VPI Bridge 1 2 Links VHDL and Verilog simulators
Windows Binaries 1 2 3 4 5 6 Detects Windows CFG bypasses and COM hijacks
Protocol Analysis
Capability Code Description
Agent Security 1 2 Audits AI endpoints and trust chains
API Security 1 Discovers APIs and tests transport weaknesses
CAN Decoding 1 Maps CAN frames using DBC metadata
Nginx Security 1 Fingerprints Nginx and probes exposed endpoints
SPECS Verifier 1 2 Verifies all 256 SPECS read counts
State Machines 1 Reconstructs ARM64 protocol state machines
TLS Analysis 1 2 Audits TLS certificates and downgrade paths
Live Ecosystems
Capability Code Description
Cisco Assessment 1 2 Audits Cisco configurations and trust paths
Industrial Protocols 1 2 Discovers protocols, credentials, and lateral movement
Kubernetes Security 1 2 Assesses container privilege and metadata risks
Malware Forensics 1 2 Collects forensic artifacts and recovers memory binaries
Privilege Assessment 1 2 Inventories hosts and privilege-escalation misconfigurations
VoIP Security 1 Enumerates VoIP services and security weaknesses
DocIN (Documentation + Integrity)
Capability Code Description
Finding Export 1 2 3 Exports findings as YARA, SARIF, and JSON
FORGE Gate 1 Blocks commits with high-risk audit findings
Ninja Plugin 1 Annotates Binary Ninja with analysis findings
Patch Tracking 1 Matches functions and reports patch changes
Sequence Matching 1 2 Matches functions and localizes patched windows
Structural Matching 1 Matches functions using control-flow topology
Supply Chain 1 2 Extracts library provenance into SBOM records
Threat Classification 1 Labels findings with MITRE ATT&CK IDs
Rubric Library
Capability Description
Index Applies versioned LLM behavior specifications to analysis tasks

Setup & About

Universal OS Installation Prompt
Ask me to choose one installation type and wait for my answer before installing anything:

1. GEM — install the standard Ablation CLI and GEM, then launch the workbench with you—the AI Agent I am currently using—embedded in its terminal.
2. CLI Only — install the standard Ablation CLI without GEM.

Do not continue until I explicitly choose GEM or CLI Only.

If I choose GEM:
- Check my operating system and Python first. Native Windows requires Python 3.11 x64 with Tcl/Tk; macOS and Linux require Python 3.10 or newer with Tkinter. Explain missing prerequisites and ask before system changes.
- Download and review https://github.com/Ablation-Tool/ablation/releases/latest/download/forall.py and run it with the checked Python. Use the default full installation, which includes both Ablation and GEM, verifies release checksums, installs GEM's wheels offline, and downloads standard Ablation dependencies from PyPI.
- Open GEM and choose the RE target inside the application; the session is created automatically outside the target. On macOS, open the generated GEM.app. On Windows and macOS, use --toolkit tk when launching through the CLI. Use your actual executable for the embedded AI Agent; if you cannot identify it, ask me instead of substituting another agent.
- Verify ablation --help and gem --help, choose a target in GEM, and confirm that the workbench opens with your real CLI inside it. Select an artifact, ask the agent to begin working on the selected item, and verify that it reads the current GEM selection and records an evidence-backed status update visible in the Lab; do not treat terminal startup alone as proof of integration.
- If I explicitly do not want a Desktop shortcut, use python forall.py --no-desktop-shortcut. Otherwise, confirm that the installed icon opens GEM directly with in-app Open binary and Open directory actions, stable per-target sessions, installed Kimi, Claude, or Codex CLI detection when available, and the GEM session and workspace environment supplied to its embedded terminal. Use gem launch --no-agent only when I explicitly want the investigation tree without an agent. Dropping local files or folders into the embedded CLI inserts quoted paths without pressing Enter; never execute an RE artifact merely to validate it.
- The turned-A logo exposes **a11y Optimization**, **OS Optimization**, and **Debug**. Each provides a read-only audit with visible findings and proposed commands before requesting permission; Deny must do nothing, and Allow may send only the displayed bounded request to the active AI agent. Accessibility includes persistent colorblind-safe and high-contrast palettes, descriptive labels, text alternatives, and keyboard navigation. Do not claim screen-reader or accessibility compliance without testing with the user's assistive technology.

If I choose CLI Only:
- Create a project-local virtual environment and activate it for my operating system. For an existing source checkout, verify its origin and preserve uncommitted changes.
- Install the standard package with python -m pip install ablation-tool, or from a reviewed source checkout with python -m pip install . Do not run forall.py because its default includes GEM.
- Verify ablation --help. Do not launch GEM.

For either choice:
- Preserve my files and local changes. Do not use sudo, overwrite an existing installation, skip failed validation, or bypass my permissions.
- Do not install an AI model, credentials, optional analysis extras, or firmware analysis. Do not execute the RE target.
- Report my selected option, operating system, Ablation and GEM versions when applicable, installation and project paths, validation result, and exact launch command.
Acknowledgments

Ablation couldn't exist without these creative individuals.

Though no code was written by these authors, doctors, and engineers, their work has direct influence on the production of code, and the aesthetics that makes Ablation what it is.

Thank you.

-Nick

Research Papers

Title Authors
Reverse Compilation Techniques · Specifying the Semantics of Machine Instructions · UQBT: Adaptable Binary Translation at Low Cost · Machine-Adaptable Dynamic Binary Translation Dr. Cristina Cifuentes
Design of a Retargetable Decompiler for a Static Platform-Independent Malware Analysis Petr Zemek, Lukáš Ďurfina, Jakub Křoustek, Dušan Kolář, Tomas Hruska, Karel Masařík, Alexander Meduna
Finding Taint-Style Vulnerabilities in Linux-based Embedded Firmware with SSE-based Alias Analysis Cheng, Zheng, Liu, Guan, Liu, Li, Zhu, Ye, Sun
iResolveX: Multi-Layered Indirect Call Resolution via Static Reasoning and Learning-Augmented Refinement Monika Santra, Bokai Zhang, Mark Lim, Vishnu Asutosh Dasu, Dongrui Zeng, Gang Tan
Extracting Protocol Format as State Machine via Controlled Static Loop Analysis Qingkai Shi, Xiangzhe Xu, Xiangyu Zhang
NEMETYL: Message Type Identification of Binary Network Protocols using Continuous Segment Similarity Stephan Kleber, Rens Wouter van der Heijden, Frank Kargl
Imperfect Forward Secrecy: How Diffie-Hellman Fails in Practice David Adrian, Karthikeyan Bhargavan, Zakir Durumeric, Pierrick Gaudry, Matthew Green, J. Alex Halderman, Nadia Heninger, Drew Springall, Emmanuel Thomé, Luke Valenta
Nonce-Disrespecting Adversaries: Practical Forgery Attacks on GCM in TLS Hanno Böck, Aaron Zauner, Sean Devlin, Juraj Somorovsky, Philipp Jovanovic
Whitening Sentence Representations for Better Semantics and Faster Retrieval Jianlin Su, Jiarun Cao, Weijie Liu, Yangyiwen Ou
Constant Propagation with Conditional Branches Mark N. Wegman, F. Kenneth Zadeck
A Simple, Fast Dominance Algorithm Cooper, Harvey, Kennedy
libdft: Practical Dynamic Data Flow Tracking for Commodity Systems Vasileios P. Kemerlis, Georgios Portokalidis, Kangkook Jee, Angelos D. Keromytis

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
Agentic GraphRAG Anthony Alcaraz, Sam Julien

GEM Design and Direction

Dr. Helen R Horkan contributed ideas and design feedback that helped shape GEM’s development and the direction of its native workbench. I’m grateful for her insight and for the influence it had on the project.

Honorable Mention

Microsoft Excel (Data Analysis ToolPak). When analyzing closed infrastructure or securing black-box systems, this A681 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.

Real-World Results

Ablation has been used to analyze production firmware and kernel drivers from Cisco, Fortinet, IBM, Juniper, 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
CVE‑2026‑20241 Identity Services Engine (ISE) and ISE Passive Identity Connector (ISE-PIC) Authorization Bypass 5.3 Medium
Demonstrations

Ablation + GEM (Graphics Environment Manager) architecture notes

These visual notes explain the naming, framework relationship, shared Lab protocol, and single Ablation + GEM (Graphics Environment Manager) distribution.

Comparison of the former GUIAI workbench with GEM (Graphics Environment Manager) and its graphical verification capabilities GEM (Graphics Environment Manager) naming and architecture layers
AI Agent interaction with the shared GEM (Graphics Environment Manager) Lab Ablation framework dependency and GEM (Graphics Environment Manager) graphical environment
One Ablation and GEM (Graphics Environment Manager) distribution with a CLI Only option

GEM (Graphics Environment Manager) Workbench

Investigation Lab and embedded AI terminal GEM settings controls and optimization buttons Accessibility audit and keyboard navigation options
OS Optimization read-only audit with permission controls Debug diagnostics with pass and review findings Node details for an investigation artifact
Resizable Lab and embedded terminal panes Full-width Lab with radial chart Treemap visualization of investigation nodes
Lab tree and embedded AI terminal layout Treemap chart beside the embedded AI terminal Radial chart beside the embedded AI terminal
Expanded embedded AI terminal workspace Workspace layout and terminal preferences Dragging a folder into the embedded AI terminal
Investigation review statuses and flagged follow-up Reviewed status filter for Lab nodes Status selection menu for Lab review
Teal accent applied to the desktop workbench Teal OS accent color in GEM Green OS accent color in GEM
Orange OS accent color in GEM Red OS accent color in GEM Purple OS accent color in GEM
Code Standards

NIST SP 800-218

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 Secure by Design

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.

MIL-HDBK-115C

The US Department of Defense handbook for reverse engineering methodology 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.
Responsible Use & Contact

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.

Contact: Feel free to contact me if you have any questions, ideas, complaints, or bugs:

About

Ablation is an AI-assisted reverse engineering (AIARE) framework

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

5 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages

0