Privacy Leakage in Language Models

Follow this topic Ask about this topic

Momentum

19 papers in the last four weeks, up 58% on the four weeks before. 0.2% of all new papers.

Jul 13Week of Sep 28

Latest papers 159

All topics

LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection
Aug 12, 2026Xinhao Zhong, Yuxia Qiao, Junhao Li +3VLM Unlearning Multimodal Large Language Models
3Pengcheng Laboratory · 1Harbin Institute of Technology, Shenzhen · 2Tsinghua University
Stealing Reasoning Traces from Proprietary LLM APIs
Aug 10, 2026Alexander Panfilov, David Schmotz, Ilia Shumailov +5Model Extraction Attacks Adversarial Attacks
MATS Research · ELLIS Institute Tübingen · Max Planck Institute for Intelligent Systems +4
Governing the KV Cache: Preventing Timing Side-Channel Leakage in Multi-Tenant LLM Inference
Aug 10, 2026Tejasvi C. AddagadaKV Caching KV-Cache Management
Beyond Direct Identifiers: Probabilistic Privacy Risk Estimation for Privacy-Conscious LLM Query Delegation
Aug 10, 2026Li Siyan, Zhou Yu, Julia HirschbergPrivacy-Preserving Language Models Privacy Leakage in Language Models
Department of Computer Science Columbia University
Mitigating Over-Personalization in LLMs via Structured Memory
Aug 8, 2026Hakeem Hannoon, Andrew Zhao, Mihir Narayan +2LLM Sycophancy Persistent Memory for Language Models
University of Saskatchewan · University of California–Irvine · University of Wisconsin–Madison +2
Protecting patient privacy in clinical foundation models: Technical and legal perspectives
Aug 7, 2026Sana Tonekaboni, Lena Stempfle, Sasha Ronaghi +4Privacy Auditing Privacy Leakage in Language Models
Massachusetts Institute of Technology (MIT), Cambridge, Massachusetts, USA · Broad Institute of MIT and Harvard, Cambridge, Massachusetts, USA · Stanford University, Stanford, California, USA +5
GRASP: Reinforcing Language Model Anonymizers with Group Relative Policy Optimization
Aug 6, 2026Sajjad Ghiasvand, Nader SehatbakhshOn-Device Language Model Inference Privacy-Preserving Language Models
Electrical and Computer Engineering Department, UC Santa Barbara1 · Electrical and Computer Engineering Department, UC Los Angeles2
DeepInvert: Semi-Supervised Embedding Inversion Against Obfuscated Language Models
Aug 5, 2026Zhicong Huang, Cheng Hong, Tao WeiPrivacy-Preserving Language Models Privacy Leakage in Language Models
Ant Group
When Agents Learn to Be You: Benchmarking Privacy Leakage, Impersonation Risk, and Defenses in Persona Skills
Aug 4, 2026Yongli Xiang, Zhifang Zhang, Bojun Yang +4Privacy Leakage in Language Models AI Agent Security Benchmarks
1Sydney AI Centre, The University of Sydney · University of Queensland · 3Southeast University +1
SparSEEty: Extracting Tokens from Sparsity-Exploiting LLM Serving Systems via Deterministic Side Channels
Aug 4, 2026Yongwan Jo, Jinyoung Park, Euihyun Lee +1Privacy Leakage in Language Models LLM Security
Department of Computer Science Yonsei University
MineGrad: Gradient Inversion Attacks on LoRA Fine-Tuning
Aug 2, 2026Hasin Us Sami, Swapneel Sen, Basak GulerGradient Inversion Attacks Privacy Leakage in Language Models
University of California, Riverside, CA
Toward Fine-Grained Forgetting:Attribute Unlearning for Multimodal Large Language Models
Aug 2, 2026Junkai Lin, Junkai Chen, Siqi Hou +5VLM Unlearning Privacy Leakage in Language Models
Institute of Automation, Chinese Academy of Sciences · The Chinese University of Hong Kong · The Australian National University +3
Leak It: Per-Document Extraction Beyond Aggregate Membership Inference
Jul 31, 2026Victor MaricatoMembership Inference Attacks LLM Auditing
Karolinska Institutet, Sweden
TriShield: Zero-Utility-Loss Defense Against Privacy Backdoors in Federated Language Model Fine-Tuning via Orthogonal Gradient Projection and Optimizer State Entanglement
Jul 30, 2026Cheng WeiBackdoor Attacks Backdoor Attacks in Federated Learning
Honor Device Co., Ltd., Shenzhen, China
Crossing the Margin Cliff: Toward Relearn-Robust LLM Unlearning via Margin Calibration
Jul 30, 2026Xiangyu Yin, Jiaxu Liu, Zhen Chen +1Privacy Leakage in Language Models LLM Unlearning
Chalmers University of Technology, Gothenburg, Sweden · Imperial College London, London, United Kingdom · University of Liverpool, Liverpool, United Kingdom +1
Subtract, Transport, or Replay? Auditable Deletion from Language-Model Memory
Jul 30, 2026Vishwajith RameshMemory-Augmented Language Models LLM Auditing
Vy Labs, Inc.
Bits and Memories: Measuring Verbatim Extraction Across LLM Quantization
Jul 28, 2026Akshay SasiLLM Quantization Memorization in Language Models
Independent Researcher
How Much Can a LoRA Adapter Memorize? Measuring Adapter Capacity in Bits
Jul 23, 2026Kaizhen Tan, Heqing Du, Yang FengMemorization in Language Models Privacy Leakage in Language Models
Carnegie Mellon University, Pittsburgh, PA, USA · Columbia University, New York, NY, USA
SlotGuard: Stop Oversharing Private Local Context in LLM Agent Transcri
Jul 19, 2026Haocheng Xia, Yongjoo ParkLLM Agent Security Privacy Leakage in Language Models
Siebel School of Computing and Data Science, University of Illinois Urbana-Champaign.
Do Speech Tokens Leak Voiceprints? Speaker Inversion Attacks Against End-to-End Speech Language Models
Jul 18, 2026Ye Lu, Yihan Yan, Zhaoyang Zhang +4Adversarial Attacks Privacy Leakage in Language Models
School of Cyberspace Science, Harbin Institute of Technology, Harbin 150001, China
PANOPTICON: A PII-Based Assemblage of Naturalistic Output Tokens for Investigating Privacy Leakage Within LLM Context Window
Jul 17, 2026Ryan Thornton, Mir Mehedi Ahsan Pritom, Maanak GuptaAdversarial Attacks Privacy Leakage in Language Models
Department of Computer Science Tennessee Tech University Cookeville, TN, USA
Privacy Leakage in Federated Learning in Radiology Reports: A Comparative Evaluation of Tokenizer-Driven Privacy Risks
Jul 15, 2026Santhosh Parampottupadam, Andres Martinez, Dimitrios Bounias +3Gradient Inversion Attacks Adversarial Attacks
German Cancer Research Center (DKFZ), Division of Medical Image Computing, Heidelberg, Germany · Medical Faculty Heidelberg, Heidelberg University, Heidelberg, Germany · Department of Computer Engineering, Bilkent University, Universiteler 06800 Çankaya/Ankara, Türkiye +1
Policy-Conditioned Constrained Decoding for Column-Level Access Control in Text-to-SQL
Jul 14, 2026Ryoto Miyamoto, Xin Fan, Hayato YamanaPrivacy Leakage in Language Models Text-to-SQL
Waseda University
Securing LLMs in the Wild: Privacy and Security Challenges at the Edge
Jul 13, 2026Ren-Yi Huang, Mingchen Li, Dumindu Samaraweera +1LLM Inference Efficiency Edge Computing
Department of Electrical and Computer Engineering, University of South Florida · Department of Mathematics, Embry-Riddle Aeronautical University
PromptGraph: Graph-Guided Prompt Sanitization for Balancing Privacy and Utility in LLM Inference
Jul 12, 2026Chen Gu, Hui Wan, Donghui Hu +2Privacy-Preserving Language Models Privacy Leakage in Language Models
School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, China
Multi-Agent Firewall Architecture for Privacy Protection of Sensitive Data in Interactions with Language Models
Jul 9, 2026Hugo García Cuesta, Pablo Mateo Torrejón, Alfonso Sánchez-MaciánPrivacy Leakage in Language Models LLM Security
Universidad Carlos III de Madrid Madrid, Spain
Overthinking: Amplifying Reasoning Weights to Extract Learned Secrets
Jul 9, 2026Jack Hopkins, Dipika Khullar, Fabien RogerOverthinking in Language Models Privacy Leakage in Language Models
Anthropic Fellows Program · MATS · Anthropic.
POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with Prompt-Optimized Parameter Shaking
Jul 7, 2026Zhangheng LI, Jianing Zhu, Junyuan Hong +4Multimodal Large Language Models Privacy Leakage in Language Models
University of Texas at Austin · DEVCOM Army Research Laboratory
Privilege and confidentiality in generative AI workflows
Jul 6, 2026Václav Janeček, Thomas MelhamData Leakage Privacy Leakage in Language Models
University of Bristol Law School · University of Oxford, Department of Computer Science
PPE-Bench: A Benchmark for Evaluating MLLM Unlearning under Private-Public Entanglement
Jul 3, 2026Xianren Zhang, Delvin Ce Zhang, Dongwon Lee +1Privacy Leakage in Language Models MLLM Unlearning
The Pennsylvania State University · University of Sheffield

Aug 12, 2026, cs.LGJ/K move · Enter open · S save

LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection

Xinhao Zhong, Yuxia Qiao, Junhao Li, Hao Fang+2

3Pengcheng Laboratory · 1Harbin Institute of Technology, Shenzhen · 2Tsinghua University

VLM Unlearning Multimodal Large Language Models Privacy Leakage in Language Models MLLM Unlearning Machine Unlearning

Open paper AskPDF arXiv

Reinforcement-learning (RL) post-training equips multimodal large reasoning models (MLRMs) with exploratory chains of thought (CoT), substantially improving visual reasoning. However, we find that this capability introduces a distinct privacy vulnerability: even when a sensitive fact is successfully unlearned from the final answer, the model may still reproduce it in its reasoning trace. This leakage is substantially more pronounced in natively RL-trained MLRMs than in their non -reasoning base models, revealing a privacy risk that existing unlearning methods are not designed to address. We show that RL-induced exploration leaves sensitive content with a distinctive token-level entropy signature that is largely absent from base models. Based on this observation, we propose LEMUR, a fully training-free, inference-time unlearning framework for natively RL-trained multimodal models. LEMUR uses entropy dynamics as a control signal to identify when sensitive reasoning begins and when sanitization should stop. During this interval, it redirects the reasoning trajectory through entropy-modulated visual-anchor latent injection, replacing committed tokens with sanitized, probability-weighted embeddings re-grounded in the input image. Across diverse MLRMs, LEMUR consistently outperforms existing unlearning met hods in suppressing both reasoning-trace and answer leakage, while better preserving non-sensitive utility and output fluency. These results demonstrate that RL-induced entropy dynamics provide a distinctive signal for privacy leakage and that exploiting this signal enables effective training-free unlearning for reasoning-capable multimodal models.

Momentum

Related topics

Latest papers 159