ML model security auditing — pickle exploit detection, integrity checks, OWASP ML Top 10
Project description
🔍 NullSec ModelAudit
ML Model Security Auditing Framework
Comprehensive security auditing for deployed machine learning models
🎯 Overview
NullSec ModelAudit is a security auditing framework for machine learning models. It inspects model files for hidden payloads (pickle deserialization, Lambda layers), checks for backdoors via Neural Cleanse and Meta Neural Analysis, evaluates robustness boundaries, and generates compliance-ready audit reports covering OWASP ML Top 10 risks.
⚡ Features
| Feature | Description |
|---|---|
| File Inspector | Detect pickle exploits, malicious Lambda layers, hidden ops |
| Backdoor Scan | Neural Cleanse, Meta Neural Analysis, fine-pruning checks |
| Robustness Eval | Automated adversarial boundary testing |
| Supply Chain | Verify model provenance and hash integrity |
| Fairness Audit | Bias detection across protected attributes |
| OWASP ML Top 10 | Map findings to OWASP ML risk categories |
| Report Engine | HTML/PDF/JSON audit reports with severity ratings |
📋 Audit Checks
| Check | Category | Severity |
|---|---|---|
| Pickle RCE | Deserialization | Critical |
| Lambda Injection | Model Architecture | Critical |
| Backdoor Trigger | Integrity | High |
| Adversarial Fragility | Robustness | High |
| Training Data Leakage | Privacy | High |
| Model Watermark | Provenance | Medium |
| Bias / Fairness | Compliance | Medium |
| Dependency Vuln | Supply Chain | Variable |
🚀 Quick Start
# Full security audit of a model file
nullsec-modelaudit scan --model model.pt --format pytorch --output audit-report.html
# Check for deserialization exploits in pickle files
nullsec-modelaudit inspect --model model.pkl --check deserialization
# Backdoor detection scan
nullsec-modelaudit backdoor --model model.h5 --dataset validation/ --num-classes 10
# Supply chain verification
nullsec-modelaudit verify --model model.onnx --expected-hash sha256:abc123...
🔗 Related Projects
| Project | Description |
|---|---|
| nullsec-adversarial | Adversarial ML attack toolkit |
| nullsec-datapoisoning | Training data poisoning detection |
| nullsec-llmred | LLM red-teaming framework |
| nullsec-promptinject | Prompt injection payloads |
| nullsec-linux | Security Linux distro (140+ tools) |
⚠️ Legal
For authorized security auditing only. Always obtain proper authorization before auditing third-party models.
📜 License
MIT License — @bad-antics
Part of the NullSec AI/ML Security Suite
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file nullsec_modelaudit-0.1.0.tar.gz.
File metadata
- Download URL: nullsec_modelaudit-0.1.0.tar.gz
- Upload date:
- Size: 5.7 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
35548a30e6237a32b7711037402e61ebbd17bd932d2378eda7ac4abbcfe645a3
|
|
| MD5 |
7080fa2c37ebc35e83015aa343af39b2
|
|
| BLAKE2b-256 |
12220b951fc8fe0ce26ea8eb6df3b7c9d91791292a2d02cbd7aed1868860479d
|
File details
Details for the file nullsec_modelaudit-0.1.0-py3-none-any.whl.
File metadata
- Download URL: nullsec_modelaudit-0.1.0-py3-none-any.whl
- Upload date:
- Size: 6.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
8779693ab1add3c74ee24eb14a0659b9802451b9e53df4b9df327f27095c363c
|
|
| MD5 |
6fcee3c4d4300d766ac7751041fe322e
|
|
| BLAKE2b-256 |
aa5920a27535ddd2a6b001c552f1ae061bbbfb27e624321fea68ff449c0b1dfd
|