跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.20454 · v1

KVMARK: Fragile Watermarking for KV-Cache Integrity in Latent Multi-Agent Systems

首次提交 2026-11-16 11:23
版本更新 2026-11-16 11:23
雷达观测 2026-11-17 13:07
arXiv ↗PDF ↗

摘要

Latent multi-agent communication directly transfers Key-Value (KV) cache tensors between transformer-based agents, bypassing text serialization entirely. While computationally efficient and reasoning accurate, this communication channel is opaque and vulnerable to man-in-the-middle tampering, a threat vector unaddressed by existing text-based or behavioral watermarks, which cannot operate on continuous internal representations. To address this vulnerability, we propose KVMark, a training-free, zero-bit KV-cache watermarking framework to verify inter-agent handoff integrity. Our framework introduces four core technical components: (1) position-dependent sign patterns that guarantee first-order statistical invisibility while preserving downstream attention survival; (2) per-head sign patterns with aligned detection to defeat head-averaging attacks; (3) a three-tier depth-aware strength schedule that equalizes detection sensitivity across transformer layers while maintaining invisibility against nine statistical profiling strategies; and (4) content-bound key derivation that cryptographically binds watermarks to input contexts, preventing replay attacks under Kerckhoffs's principle. Empirical evaluation on a four-agent latent pipeline demonstrates 100%100\% detection across 1,5001{,}500 handoffs with a 0.0%0.0\% false positive rate (AUROC=1.0\text{AUROC}=1.0), 100%100\% layer-level tamper localization (17/1717/17 attacked layers identified with 0/190/19 false positives), and no statistically significant reasoning accuracy degradation across four benchmarks (N=300N{=}300--500500). Furthermore, the framework enables multi-hop cascade verification to isolate compromised agents (500/500 detection at three hops), withstands adaptive white-box adversaries and the Out-of-Sight DiffMean attack (up to α=8.0α=8.0), and generalizes seamlessly to Qwen3-8B without re-tuning (1,500/1,500 TPR).

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-16 11:23 · 最新观测版本

KVMARK: Fragile Watermarking for KV-Cache Integrity in Latent Multi-Agent Systems

Latent multi-agent communication directly transfers Key-Value (KV) cache tensors between transformer-based agents, bypassing text serialization entirely. While computationally efficient and reasoning accurate, this communication channel is opaque and vulnerable to man-in-the-middle tampering, a threat vector unaddressed by existing text-based or behavioral watermarks, which cannot operate on continuous internal representations. To address this vulnerability, we propose KVMark, a training-free, zero-bit KV-cache watermarking framework to verify inter-agent handoff integrity. Our framework introduces four core technical components: (1) position-dependent sign patterns that guarantee first-order statistical invisibility while preserving downstream attention survival; (2) per-head sign patterns with aligned detection to defeat head-averaging attacks; (3) a three-tier depth-aware strength schedule that equalizes detection sensitivity across transformer layers while maintaining invisibility against nine statistical profiling strategies; and (4) content-bound key derivation that cryptographically binds watermarks to input contexts, preventing replay attacks under Kerckhoffs's principle. Empirical evaluation on a four-agent latent pipeline demonstrates 100%100\% detection across 1,5001{,}500 handoffs with a 0.0%0.0\% false positive rate (AUROC=1.0\text{AUROC}=1.0), 100%100\% layer-level tamper localization (17/1717/17 attacked layers identified with 0/190/19 false positives), and no statistically significant reasoning accuracy degradation across four benchmarks (N=300N{=}300--500500). Furthermore, the framework enables multi-hop cascade verification to isolate compromised agents (500/500 detection at three hops), withstands adaptive white-box adversaries and the Out-of-Sight DiffMean attack (up to α=8.0α=8.0), and generalizes seamlessly to Qwen3-8B without re-tuning (1,500/1,500 TPR).

查看此版本解读 →查看此版本原文 ↗