When a prompt injection attack succeeds, a Large Language Model (LLM) abandons its assigned system role to comply with an adversarial instruction. While prior work has extensively quantified how often this occurs, we ask a more fundamental question: where inside the network does the model actually decide to break the r...
Rui Wen, Jia-Yang Liu, Ze-Yu Yang et al.· 0 citations
Across vision and language models, it is shown that efficiency-oriented training increases susceptibility to adversarial and privacy attacks, and is called for a paradigm shift toward multi-objective training that jointly optimizes for performance, cost, and security.
Yi-Yong Liu, Jun Sakuma, Michael Backes et al.· 0 citations
VIBMask is proposed, a unified IWFS framework by the variational information bottleneck that mitigates co-adaptation by penalizing mutual information between unselected features and the label, and improves expressivity via an ensemble of diverse selectors that capture heterogeneous sparse patterns.
Lu Sun, Jun Sakuma· Proceedings of the Thirty-Fi...· 0 citations
Neutral Prompting Attack (NPA), a highly stealthy attack paradigm in which semantically benign instructions, such as encouraging imagination and exhaustiveness, increase package hallucination propensity without containing explicit malicious intent is introduced.
