SURA detects federated-learning backdoors with 10 clean images
It reuses one small clean proxy across repeated loss-ascent probes and restores batch-norm state only at readout.
Stories per day · 30 days
1 RSS
Relatedbackdoor2ai-safety1batch-normalization1codex1federated-learning1open-weights1supply-chain1unlearning1
8 stories, newest first · one card per story
Stories tagged #model-poisoning
Backdooring Open-Weight AI Models
Researchers showed open-weight models can hide backdoors that pass evals and exfiltrate credentials when a trigger appears.
13
Byzantine-Resilient Federated Large Language Model via Hypergraph Message Passing
HMP-GAE filters Byzantine updates during federated training of Qwen-2.5 and Llama-3.2.
Compression Footprints as Security Signals for Model-Poisoning Defense in Federated Learning
CRAFT defends federated learning against model poisoning by scoring lossy-compression footprints of client updates.
Cybersecurity in Edge Computing: A Trust-Aware Federated Hybrid Intrusion Detection Framework
Researchers propose a trust-aware federated hybrid IDS that detects edge attacks without sharing raw telemetry.
OPFL: Optimistic Verification of Federated Learning via Empirical Boundary
OPFL verifies federated learning inside MPC with an empirical gradient boundary, stopping model poisoning at much lower cost.
Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation
NEEDLE removes LLM backdoors by weight orthogonalisation with little capability or safety loss.
When the World Lies: Backdoor Attacks on Latent World Models for Downstream Control
A poisoned world-model checkpoint hijacks downstream controllers without an explicit trigger rule, passing clean-data evaluation while steering 100% of triggered actions.