Cover Image

Observable Channels, Not Just Storage: Evaluating Privacy Leakage in LLM Agent Pipelines

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Huang, Tao, Hou, Chen, Wu, Guosen, Meng, Jiayang
Format:	Preprint
Published:	2026
Subjects:	Cryptography and Security
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Real-Time Privacy Risk Measurement with Privacy Tokens for Gradient Leakage
by: Meng, Jiayang, et al.
Published: (2025)

Enhanced Privacy Leakage from Noise-Perturbed Gradients via Gradient-Guided Conditional Diffusion Models
by: Meng, Jiayang, et al.
Published: (2025)

Is Diffusion Model Safe? Severe Data Leakage via Gradient-Guided Diffusion Model
by: Meng, Jiayang, et al.
Published: (2024)

Rethinking Side-Channel Analysis: Automated Discovery and Analysis of Side-Channel Leakage with LLM-Assisted Agents
by: Xu, Zhen, et al.
Published: (2026)

RTBAS: Defending LLM Agents Against Prompt Injection and Privacy Leakage
by: Zhong, Peter Yong, et al.
Published: (2025)

LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage
by: Nie, Yuzhou, et al.
Published: (2024)

Secure Distributed Storage: Optimal Trade-Off Between Storage Rate and Privacy Leakage
by: Chou, Remi A., et al.
Published: (2024)

Beyond TVLA: Anderson-Darling Leakage Assessment for Neural Network Side-Channel Leakage Detection
by: Mikulec, Ján, et al.
Published: (2026)

PrivacyPeek: Auditing What LLM-Based Agents Acquire, Not Just What They Say
by: Zhang, Mingxuan, et al.
Published: (2026)

Computing Maximal Per-Record Leakage and Leakage-Distortion Functions for Privacy Mechanisms under Entropy-Constrained Adversaries
by: Wu, Genqiang, et al.
Published: (2026)

PriRoAgg: Achieving Robust Model Aggregation with Minimum Privacy Leakage for Federated Learning
by: Hou, Sizai, et al.
Published: (2024)

Exposing LLM User Privacy via Traffic Fingerprint Analysis: A Study of Privacy Risks in LLM Agent Interactions
by: Zhang, Yixiang, et al.
Published: (2025)

SMOTE and Mirrors: Exposing Privacy Leakage from Synthetic Minority Oversampling
by: Ganev, Georgi, et al.
Published: (2025)

Evaluating Privacy Leakage in Split Learning
by: Qiu, Xinchi, et al.
Published: (2023)

Profiling for Pennies: Unveiling the Privacy Iceberg of LLM Agents
by: Chen, Jiahao, et al.
Published: (2026)

The Inevitability of Side-Channel Leakage in Encrypted Traffic
by: Liu, Guangjie, et al.
Published: (2026)

Evaluating Differential Privacy on Correlated Datasets Using Pointwise Maximal Leakage
by: Saeidian, Sara, et al.
Published: (2025)

Privacy in Action: Towards Realistic Privacy Mitigation and Evaluation for LLM-Powered Agents
by: Wang, Shouju, et al.
Published: (2025)

CRFU: Compressive Representation Forgetting Against Privacy Leakage on Machine Unlearning
by: Wang, Weiqi, et al.
Published: (2025)

S-Leak: Leakage-Abuse Attack Against Efficient Conjunctive SSE via s-term Leakage
by: Su, Yue, et al.
Published: (2025)

Zebrafix: Mitigating Memory-Centric Side-Channel Leakage via Interleaving
by: Pätschke, Anna, et al.
Published: (2025)

POLARIS: Explainable Artificial Intelligence for Mitigating Power Side-Channel Leakage
by: Mahfuz, Tanzim, et al.
Published: (2025)

Revisiting Privacy Leakage in Machine Unlearning: Membership Inference Beyond the Forgotten Set
by: Fu, Jie, et al.
Published: (2026)

PrivTru: A Privacy-by-Design Data Trustee Minimizing Information Leakage
by: Gehring, Lukas, et al.
Published: (2025)

Credential Leakage in LLM Agent Skills: A Large-Scale Empirical Study
by: Chen, Zhihao, et al.
Published: (2026)

Causality Laundering: Denial-Feedback Leakage in Tool-Calling LLM Agents
by: Chinaei, Mohammad Hossein
Published: (2026)

The Hidden Cost of Correlation: Rethinking Privacy Leakage in Local Differential Privacy
by: Jayawardana, Sandaru, et al.
Published: (2025)

The Early Bird Catches the Leak: Unveiling Timing Side Channels in LLM Serving Systems
by: Song, Linke, et al.
Published: (2024)

Metric-Normalized Posterior Leakage (mPL): Attacker-Aligned Privacy for Joint Consumption
by: Chen, Gaoyi, et al.
Published: (2026)

Quantifying Privacy Leakage in Split Inference via Fisher-Approximated Shannon Information Analysis
by: Deng, Ruijun, et al.
Published: (2025)

Intermediate Outputs Are More Sensitive Than You Think
by: Huang, Tao, et al.
Published: (2024)

From Storage to Steering: Memory Control Flow Attacks on LLM Agents
by: Xu, Zhenlin, et al.
Published: (2026)

HarnessAgent: Scaling Automatic Fuzzing Harness Construction with Tool-Augmented LLM Pipelines
by: Yang, Kang, et al.
Published: (2025)

Control at Stake: Evaluating the Security Landscape of LLM-Driven Email Agents
by: Wu, Jiangrong, et al.
Published: (2025)

The Emerged Security and Privacy of LLM Agent: A Survey with Case Studies
by: He, Feng, et al.
Published: (2024)

Membership Information Leakage in Federated Contrastive Learning
by: Chen, Kongyang, et al.
Published: (2024)

Analysis of Privacy Leakage in Federated Large Language Models
by: Vu, Minh N., et al.
Published: (2024)

A Framework for Managing Multifaceted Privacy Leakage While Optimizing Utility in Continuous LBS Interactions
by: Bkakria, Anis, et al.
Published: (2024)

EdgeLeakage: Membership Information Leakage in Distributed Edge Intelligence Systems
by: Chen, Kongyang, et al.
Published: (2024)

SafeGPT: Preventing Data Leakage and Unethical Outputs in Enterprise LLM Use
by: Desai, Pratyush, et al.
Published: (2026)