Pinned Loading

1. SORRY-Bench/sorry-bench SORRY-Bench/sorry-bench Public

   Benchmark evaluation code for "SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal" (ICLR 2025)

   Jupyter Notebook 50 2

2. LLM-Tuning-Safety/LLMs-Finetuning-Safety LLM-Tuning-Safety/LLMs-Finetuning-Safety Public

   We jailbreak GPT-3.5 Turbo’s safety guardrails by fine-tuning it on only 10 adversarially designed examples, at a cost of less than $0.20 via OpenAI’s APIs.

   Python 287 32

3. backdoor-toolbox backdoor-toolbox Public

   A compact toolbox for backdoor attacks and defenses.

   Python 169 21

4. Unispac/Subnet-Replacement-Attack Unispac/Subnet-Replacement-Attack Public

   Official implementation of (CVPR 2022 Oral) Towards Practical Deployment-Stage Backdoor Attack on Deep Neural Networks.

   Jupyter Notebook 26 7

5. Unispac/Fight-Poison-With-Poison Unispac/Fight-Poison-With-Poison Public

   Code repository for the paper --- [USENIX Security 2023] Towards A Proactive ML Approach for Detecting Backdoor Poison Samples

   Python 25 2

6. ain-soph/trojanzoo ain-soph/trojanzoo Public

   TrojanZoo provides a universal pytorch platform to conduct security researches (especially backdoor attacks/defenses) of image classification in deep learning.

   Python 290 64