tldrsec / prompt-injection-defenses Goto Github PK

https://arxiv.org/pdf/2312.10766

we propose JailGuard, a universal detection framework for jailbreaking and hijacking
attacks across LLMs and MLLMs. JailGuard operates on the principle that attacks are inherently less robust
than benign ones, regardless of method or modality. Specifically, JailGuard mutates untrusted inputs to
generate variants and leverages discrepancy of the variants’ responses on the model to distinguish attack
samples from benign samples