3 papers
cs.LG2026
Why Backdooring Neural Networks is so Easy?
Issam Seddik, Mohamed El Amine Seddik
Securing modern AI systems against backdoor attacks remains an open challenge and requires fundamentally principled estimates of the adversary's budget -- the poison fraction a…
cs.CR2026
The Power of Backdoor Absorption in Community Training
Issam Seddik, Sami Souihi, Mohamed Tamaazousti +1
Backdoor attacks severely threaten large-scale AI models. When model owners delegate training to external compute providers within a decentralized training paradigm, adversaries ca…
cs.CR2025
PoTS: Proof-of-Training-Steps for Backdoor Detection in Large Language Models
Issam Seddik, Sami Souihi, Mohamed Tamaazousti +1
As Large Language Models (LLMs) gain traction across critical domains, ensuring secure and trustworthy training processes has become a major concern. Backdoor attacks, where malici…