Shadow in the Cache: Unveiling and Mitigating Privacy Risks of KV-cache in LLM Inference

The Key-Value (KV) cache, which stores intermediate attention computations (Key and Value pairs) to avoid redundant calculations, is a fundamental mechanism for accelerating Large Language Model (LLM) inference. However, this efficiency optimization introduces significant yet underexplored privacy risks. This paper provides the first comprehensive analysis of these vulnerabilities, demonstrating that an attacker can reconstruct sensitive user inputs directly from the KV-cache. We design and implement three distinct attack vectors: a direct Inversion Attack, a more broadly applicable and potent Collision Attack, and a semantic-based Injection Attack. These methods demonstrate the practicality and severity of KV-cache privacy leakage issues. To mitigate this, we propose KV-Cloak, a novel, lightweight, and efficient defense mechanism. KV-Cloak uses a reversible matrix-based obfuscation scheme, combined with operator fusion, to secure the KV-cache. Our extensive experiments show that KV-Cloak effectively thwarts all proposed attacks, reducing reconstruction quality to random noise. Crucially, it achieves this robust security with virtually no degradation in model accuracy and minimal performance overhead, offering a practical solution for trustworthy LLM deployment.

Key Contributions

• First comprehensive study of KV-cache privacy risks: three novel attack vectors (Inversion Attack using model weights, Collision Attack via forward-pass matching, semantic Injection Attack) that reconstruct private user inputs from KV-cache
• KV-Cloak: a lightweight reversible matrix-based obfuscation scheme with operator fusion that secures KV-cache against all three attack classes
• Empirical demonstration that KV-Cloak reduces reconstruction quality to random noise with negligible model accuracy degradation and minimal latency overhead

🛡️ Threat Analysis

Details

Similar Papers