nazmul-karim170/FIP
[CCS'24] Official Implementation of "Fisher Information guided Purification against Backdoor Attacks"
This project helps machine learning engineers and AI security researchers to identify and remove 'backdoor' vulnerabilities from their trained AI models. It takes a suspicious trained model, analyzes its internal workings using Fisher Information, and then 'purifies' it to remove malicious behaviors, producing a safer, more reliable model. This is especially useful for those deploying AI in sensitive applications like image classification, action recognition, or natural language processing.
Use this if you need to clean a trained AI model that might have been compromised by a backdoor attack, ensuring it performs reliably without hidden malicious functions.
Not ideal if you are looking to prevent backdoor attacks during the initial training phase or need to detect if a model *could* be backdoored before it's trained.
Stars
14
Forks
2
Language
Python
License
MIT
Category
Last pushed
Oct 29, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/nazmul-karim170/FIP"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.