SewoongLab/spectre-defense
Defending Against Backdoor Attacks Using Robust Covariance Estimation
This project helps machine learning researchers and security analysts protect image classification models from 'backdoor' attacks. When a model is trained on a poisoned dataset, malicious patterns can be hidden that trigger incorrect predictions later. This tool takes a trained, potentially backdoored image classifier and its hidden data representations, then identifies and removes the poisoned samples so the model can be retrained securely.
No commits in the last 6 months.
Use this if you are a machine learning security researcher or practitioner working with image classification models that might have been compromised by data poisoning or backdoor attacks.
Not ideal if you are not working with image data, do not have access to the hidden representations of your neural network, or are looking for defenses against types of attacks other than backdoors.
Stars
22
Forks
7
Language
Python
License
MIT
Category
Last pushed
Jul 12, 2021
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/SewoongLab/spectre-defense"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
QData/TextAttack
TextAttack 🐙 is a Python framework for adversarial attacks, data augmentation, and model...
ebagdasa/backdoors101
Backdoors Framework for Deep Learning and Federated Learning. A light-weight tool to conduct...
THUYimingLi/backdoor-learning-resources
A list of backdoor learning resources
zhangzp9970/MIA
Unofficial pytorch implementation of paper: Model Inversion Attacks that Exploit Confidence...
LukasStruppek/Plug-and-Play-Attacks
[ICML 2022 / ICLR 2024] Source code for our papers "Plug & Play Attacks: Towards Robust and...