Low-Loss Subspace Compression for Clean Gains against Multi-Agent Backdoor Attacks. (arXiv:2203.03692v1 [cs.LG])

Recent exploration of the multi-agent backdoor attack demonstrated the
backfiring effect, a natural defense against backdoor attacks where backdoored
inputs are randomly classified. This yields a side-effect of low accuracy
w.r.t. clean labels, which motivates this paper's work on the construction of
multi-agent backdoor defenses that maximize accuracy w.r.t. clean labels and
minimize that of poison labels. Founded upon agent dynamics and low-loss
subspace construction, we contribute three defenses that yield improved
multi-agent backdoor robustness.

agent attacks backdoor compression lg loss