DRIP aims to achieve a new security-utility balance by “de-instructionalization”, i.e., precisely removing the instruction semantics from tokens in the data section while preserving their data semantics and robustly maintain the effectiveness of the intended instruction, even against strong adversarial content.
AEGIS (Adaptive Ensemble Guard for Injection Shielding) extracts instruction-sensitive projectors to identify malicious instructions and leverages a Unified Multi-Layer Consensus mechanism that aggregates topologically distinct signals across the network depth.
Jia-Hao Chen, Ruiping Yin, Xin-Feng Li et al.· 1 citation· ⚡1
Tool-using language-model agents are vulnerable to indirect prompt injection because they must act on untrusted external content. Existing training-time defenses can reduce attack success rates, but often at the cost of general capabilities. We show that training-based defenses induce substantial drift in the model's o...
Mohamed Dhouib, Clément Elliker, Alexi Canesse et al.· 0 citations
An exploratory case study that applies explainable artificial intelligence techniques to analyze how Prompt Guard 2 distinguishes malicious from benign prompts finds that Prompt Guard 2's decisions rely on the cumulative contribution of many tokens rather than a few dominant ones, yet saliency-guided synonym substituti...
Fernando Outeda, Gustavo Betarte, J. Campo et al.· 0 citations
Coding agents have become increasingly long-horizon, autonomous, reliant on general-purpose shell and maintain their own persistent memory for self-improvement. While these capabilities have made the agents powerful, they have also made them harder to defend against external adversaries. Defenses that restrict this arc...
Mayank Rathee, Alexander Stepanov, Shalin Madabhavi et al.· 0 citations
This paper studies the role played by the adversarial data's input modality, and identifies a systematic asymmetry: multimodal LLMs are more likely to follow adversarial instruction when they appear as text than when the same instruction is delivered through a non-textual channel.
Jie Zhang, Andrei Baroian, J. V. van Rijn et al.· 0 citations
Two universal tool-based defenses are introduced: Attacker Tool Filtering, which uses anomaly detection to identify and remove suspicious tools, and Normal Tool Recalling, a white-box method that restores the agent's original toolset prior to planning.
Xiao-Yan Li, Yun-Li Wang· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
