Backdoors Framework for Deep Learning and Federated Learning. A light-weight tool to conduct your research on backdoors.

security machine-learning research pytorch adversarial backdoors adversarial-machine-learning federated-learning backdoor-attacks neural-trojan deep-learning-security ml-backdoors deep-learning-backdoors

• Updated Feb 5, 2023
• Python

pretrained BERT model for cyber security text, learned CyberSecurity Knowledge

nlp security apt transformers cybersecurity attention threat-hunting deeplearning cyber-security nlp-machine-learning bert threat-analysis security-automation threat-intelligence machine-learning-security cyber-threat-intelligence threat-detection transformer-encoder bert-embeddings deep-learning-security

• Updated Apr 28, 2023
• Python

WaNet - Imperceptible Warping-based Backdoor Attack (ICLR 2021)

security machine-learning computer-vision deep-learning backdoor-attacks deep-learning-security iclr2021

• Updated Nov 11, 2024
• Python

Universal Adversarial Perturbations (UAPs) for PyTorch

deep-neural-networks pytorch image-classification adversarial-machine-learning adversarial-examples universal-adversarial-perturbations adversarial-attacks deep-learning-security universal-attacks

• Updated Aug 28, 2021
• Jupyter Notebook

Input-aware Dynamic Backdoor Attack (NeurIPS 2020)

security deep-learning attack neurips-2020 backdoor-attack deep-learning-security

• Updated Jul 22, 2024
• Python

COMBAT: Alternated Training for Effective Clean-Label Backdoor Attack (AAAI 2024)

machine-learning computer-vision backdoor-attacks deep-learning-security trustworthy-ai aaai2024

• Updated Aug 26, 2024
• Python

Inspired by dynamic taint tracking, PoisonSpot uses fine-grained training provenance tracker that: (1) tags & traces the impact of every single training sample on model updates, (2) probabilistically scores suspect samples based on their lineage of impact on model weights, and (3) separates the clean from the poisonous before retraining a model.

adversarial-machine-learning machine-learning-security backdoor-attacks poisoning-attack aisec deep-learning-security backdoor-defense provenance-tracking backdoor-detection poisoning-defenses training-provenance-tracking

• Updated Jan 8, 2026
• Python

Neural Network Model Reverse Engineering Toolkit

neural-network-analysis deep-learning-security model-reverse-engineering ml-model-extraction ai-model-auditing black-box-analysis model-architecture-detection neural-network-forensics ai-security-tools multi-framework-analysis

• Updated Aug 28, 2025
• Jupyter Notebook

Research on developing and testing defenses to fortify neural networks against One-Pixel Attacks.

adversarial-attacks adversarial-defense one-pixel-attack deep-learning-security neural-network-robustness

• Updated Oct 25, 2025
• Jupyter Notebook

High-performance C++20 & LibTorch library for accelerated adversarial ML attack generation. Features fused CUDA kernels for PGD and FGSM, delivering up to 22x speedup over Python Torchattacks.

cuda cuda-kernels adversarial-machine-learning cpp20 adversarial-attacks ai-security libtorch pytorch-cpp deep-learning-security pgd-attack

• Updated Aug 11, 2026
• TeX