Some New AI Trojan Papers!

[taylor.kul...@iarpa.gov]

Hello TrojAI Community,

Here are some recent papers relevant to backdoor attacks that I thought I'd share.

RAB: Provable Robustness Against Backdoor Attacks

One Sentence Summary: Develops a framework for certifying DNNs against backdoor attacks using gradient smoothing and then uses this to provide the first training procedure that can defend against backdoors

Witches' Brew: Industrial Scale Data Poisoning via Gradient Matching

One Sentence Summary:  Analyzes a data poisoning attack that is clean-label, works on randomly initialized networks, and is imperceptible to humans

I have kept adding papers like these to the TrojAI Literature GitHub page as well. Feel free to take a look there as well as post any other papers you have found!

Best,

Taylor Kulp-McDowall

Tech SETA, TrojAI

taylor.kul...@iarpa.gov