Cover Image

When Routine Chats Turn Toxic: Unintended Long-Term State Poisoning in Personalized Agents

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Xu, Xiaoyu, Du, Minxin, Xie, Qipeng, Ke, Haobin, Ye, Qingqing, Hu, Haibo
Format:	Preprint
Published:	2026
Subjects:	Cryptography and Security Computation and Language Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

OBLIVIATE: Robust and Practical Machine Unlearning for Large Language Models
by: Xu, Xiaoyu, et al.
Published: (2025)

Data Poisoning Attacks to Local Differential Privacy Protocols for Graphs
by: He, Xi, et al.
Published: (2024)

LDPRecover: Recovering Frequencies from Poisoning Attacks against Local Differential Privacy
by: Sun, Xinyue, et al.
Published: (2024)

Privacy for Free: Leveraging Local Differential Privacy Perturbed Data from Multiple Services
by: Du, Rong, et al.
Published: (2025)

Unlearning Isn't Deletion: Investigating Reversibility of Machine Unlearning in LLMs
by: Xu, Xiaoyu, et al.
Published: (2025)

Interactive Trimming against Evasive Online Data Manipulation Attacks: A Game-Theoretic Approach
by: Fu, Yue, et al.
Published: (2024)

FIT to Forget: Robust Continual Unlearning for Large Language Models
by: Xu, Xiaoyu, et al.
Published: (2026)

Federated Heavy Hitter Analytics with Local Differential Privacy
by: Zhang, Yuemin, et al.
Published: (2024)

From Domains to Instances: Dual-Granularity Data Synthesis for LLM Unlearning
by: Xu, Xiaoyu, et al.
Published: (2026)

Virus Infection Attack on LLMs: Your Poisoning Can Spread "VIA" Synthetic Data
by: Liang, Zi, et al.
Published: (2025)

Multi-class Item Mining under Local Differential Privacy
by: Mao, Yulian, et al.
Published: (2025)

PriPL-Tree: Accurate Range Query for Arbitrary Distribution under Local Differential Privacy
by: Wang, Leixia, et al.
Published: (2024)

PrivDPR: Synthetic Graph Publishing with Deep PageRank under Differential Privacy
by: Zhang, Sen, et al.
Published: (2025)

PrivShape: Extracting Shapes in Time Series under User-Level Local Differential Privacy
by: Mao, Yulian, et al.
Published: (2024)

WiFinger: Fingerprinting Noisy IoT Event Traffic Using Packet-level Sequence Matching
by: Li, Ronghua, et al.
Published: (2025)

MER-Inspector: Assessing model extraction risks from an attack-agnostic perspective
by: Zhang, Xinwei, et al.
Published: (2025)

Dual Utilization of Perturbation for Stream Data Publication under Local Differential Privacy
by: Du, Rong, et al.
Published: (2025)

AdvSGM: Differentially Private Graph Learning via Adversarial Skip-gram Model
by: Zhang, Sen, et al.
Published: (2025)

United We Defend: Collaborative Membership Inference Defenses in Federated Learning
by: Bai, Li, et al.
Published: (2026)

Membership Inference Attacks and Defenses in Federated Learning: A Survey
by: Bai, Li, et al.
Published: (2024)

CSC: Turning the Adversary's Poison against Itself
by: Shi, Yuchen, et al.
Published: (2026)

Keystroke Detection by Exploiting Unintended RF Emission from Repaired USB Keyboards
by: Bari, Md Faizul, et al.
Published: (2025)

Why Are My Prompts Leaked? Unraveling Prompt Extraction Threats in Customized Large Language Models
by: Liang, Zi, et al.
Published: (2024)

How Much Do Large Language Model Cheat on Evaluation? Benchmarking Overestimation under the One-Time-Pad-Based Framework
by: Liang, Zi, et al.
Published: (2025)

Toward Efficient Inference Attacks: Shadow Model Sharing via Mixture-of-Experts
by: Bai, Li, et al.
Published: (2025)

From Randomized Response to Randomized Index: Answering Subset Counting Queries with Local Differential Privacy
by: Ye, Qingqing, et al.
Published: (2025)

Confused ChatGPT: Cross-App Context Poisoning via First-Party APIs
by: Wang, Chao, et al.
Published: (2026)

TRUSTDESC: Preventing Tool Poisoning in LLM Applications via Trusted Description Generation
by: Ye, Hengkai, et al.
Published: (2026)

Turning Generative Models Degenerate: The Power of Data Poisoning Attacks
by: Jiang, Shuli, et al.
Published: (2024)

When Benign Inputs Lead to Severe Harms: Eliciting Unsafe Unintended Behaviors of Computer-Use Agents
by: Jones, Jaylen, et al.
Published: (2026)

CBPF: Filtering Poisoned Data Based on Composite Backdoor Attack
by: Xia, Hanfeng, et al.
Published: (2024)

MindGuard: Intrinsic Decision Inspection for Securing LLM Agents Against Metadata Poisoning
by: Wang, Zhiqiang, et al.
Published: (2025)

Improving Google A2A Protocol: Protecting Sensitive Data and Mitigating Unintended Harms in Multi-Agent Systems
by: Louck, Yedidel, et al.
Published: (2025)

When the Manual Lies: A Realistic Benchmark to Evaluate MCP Poisoning Attacks for LLM Agents
by: Liu, Shi, et al.
Published: (2026)

From Defender to Devil? Unintended Risk Interactions Induced by LLM Defenses
by: Meng, Xiangtao, et al.
Published: (2025)

Can a Single Message Paralyze the AI Infrastructure? The Rise of AbO-DDoS Attacks through Targeted Mobius Injection
by: Liang, Zi, et al.
Published: (2026)

Deep-Research Agents Can Be Poisoned via User-Generated Content
by: Zhang, Tingwei, et al.
Published: (2026)

When Secure Aggregation Falls Short: Achieving Long-Term Privacy in Asynchronous Federated Learning for LEO Satellite Networks
by: Elmahallawy, Mohamed, et al.
Published: (2025)

A Sample-Level Evaluation and Generative Framework for Model Inversion Attacks
by: Li, Haoyang, et al.
Published: (2025)

When and Where do Data Poisons Attack Textual Inversion?
by: Styborski, Jeremy, et al.
Published: (2025)