大模型对齐与安全 - arXivDaily 专题

2006.01412 2022-07-19 eess.SP cs.IT cs.LG math.IT 57%

Federated Learning in Vehicular Networks

Ahmet M. Elbir, Burak Soner, Sinem Coleri, Deniz Gunduz, Mehdi Bennis

专题命中隐私与版权：safety(abstract)；分类 cs.LG

Comments 2022 IEEE International Mediterranean Conference on Communications and Networking (MeditCom)

2204.05856 2022-04-13 stat.ML cs.CR cs.LG 57%

Distributed learning optimisation of Cox models can leak patient data: Risks and solutions

Carsten Brink, Christian Rønn Hansen, Matthew Field, Gareth Price, David Thwaites, Nis Sarup, Uffe Bernchou, Lois Holloway

专题命中隐私与版权：safety(abstract)；分类 cs.LG

Comments 51 pages, 8 figures

2202.11136 2022-02-24 cs.SD cs.LG eess.AS 57%

FlowSense: Monitoring Airflow in Building Ventilation Systems Using Audio Sensing

Bhawana Chhaglani, Camellia Zakaria, Adam Lechowicz, Prashant Shenoy, Jeremy Gummeson

专题命中隐私与版权：safety(abstract)；分类 cs.LG

Comments 26 pages, 12 figures, Will appear in March issue of the IMWUT 2022 journal

2201.07711 2022-01-20 cs.CR cs.HC cs.LG cs.OS 57%

Enhancing the Security & Privacy of Wearable Brain-Computer Interfaces

Zahra Tarkhani, Lorena Qendro, Malachy O'Connor Brown, Oscar Hill, Cecilia Mascolo, Anil Madhavapeddy

专题命中隐私与版权：safety(abstract)；分类 cs.LG

2112.06006 2021-12-23 cs.DC cs.CY 57%

Towards the Internet of Behaviors in airports with a fog-to-cloud approach

Antonio Salis

专题命中隐私与版权：safety(abstract)；分类 cs.CY

Comments 16 pages, 10 figures;

2111.14838 2021-12-01 cs.LG cs.CR 57%

Evaluating Privacy-Preserving Machine Learning in Critical Infrastructures: A Case Study on Time-Series Classification

Dominique Mercier, Adriano Lucieri, Mohsin Munir, Andreas Dengel, Sheraz Ahmed

专题命中隐私与版权：safety(abstract)；分类 cs.LG

Comments 9 pages, 4 figures. 6 tables

2108.11368 2021-08-26 cs.CV cs.LG 57%

CDCGen: Cross-Domain Conditional Generation via Normalizing Flows and Adversarial Training

Hari Prasanna Das, Ryan Tran, Japjot Singh, Yu-Wen Lin, Costas J. Spanos

专题命中隐私与版权：alignment(abstract)；分类 cs.LG

Comments Workshop on Machine Learning for Data: Automated Creation,Privacy, Bias, In 38th International Conference on Machine Learning (ICML) 2021

2106.08299 2021-06-16 cs.LG 57%

Model Extraction and Adversarial Attacks on Neural Networks using Switching Power Information

Tommy Li, Cory Merkel

专题命中隐私与版权：safety(abstract)；分类 cs.LG

2012.00138 2021-04-29 cs.LG cs.SY eess.SY 57%

Robust error bounds for quantised and pruned neural networks

Jiaqi Li, Ross Drummond, Stephen R. Duncan

专题命中隐私与版权：safety(abstract)；分类 cs.LG

1901.08755 2021-04-08 cs.LG stat.ML 57%

SecureBoost: A Lossless Federated Learning Framework

Kewei Cheng, Tao Fan, Yilun Jin, Yang Liu, Tianjian Chen, Dimitrios Papadopoulos, Qiang Yang

专题命中隐私与版权：alignment(abstract)；分类 cs.LG

1907.00164 2021-02-08 cs.LG stat.ML 57%

On the Privacy Risks of Model Explanations

Reza Shokri, Martin Strobel, Yair Zick

专题命中隐私与版权：trustworthy(abstract)；分类 cs.LG

Comments 19 pages, 13 figures

2011.07555 2020-11-17 cs.CR cs.CY 57%

Towards Compliant Data Management Systems for Healthcare ML

Goutham Ramakrishnan, Aditya Nori, Hannah Murfet, Pashmina Cameron

专题命中隐私与版权：trustworthy(abstract)；分类 cs.CY

2004.07427 2020-06-12 cs.CR cs.LG 57%

Asymmetrical Vertical Federated Learning

Yang Liu, Xiong Zhang, Libin Wang

专题命中隐私与版权：alignment(abstract)；分类 cs.LG

2001.06202 2020-01-20 cs.LG cs.CV stat.ML 57%

FedVision: An Online Visual Object Detection Platform Powered by Federated Learning

Yang Liu, Anbu Huang, Yun Luo, He Huang, Youzhi Liu, Yuanyuan Chen, Lican Feng, Tianjian Chen, Han Yu, Qiang Yang

专题命中隐私与版权：safety(abstract)；分类 cs.LG

1902.08934 2019-11-06 cs.LG cs.CR 57%

Privacy Preserving Location Data Publishing: A Machine Learning Approach

Sina Shaham, Ming Ding, Bo Liu, Shuping Dang, Zihuai Lin, Jun Li

专题命中隐私与版权：alignment(abstract)；分类 cs.LG

1812.11494 2019-01-17 cs.IT cs.LG math.IT 57%

Broadband Analog Aggregation for Low-Latency Federated Edge Learning (Extended Version)

Guangxu Zhu, Yong Wang, Kaibin Huang

专题命中隐私与版权：alignment(abstract)；分类 cs.LG

Comments This is an extended version of a submission to IEEE journal

1805.12024 2018-05-31 cs.LG cs.CV cs.NE stat.ML 57%

Privacy Aware Offloading of Deep Neural Networks

Sam Leroux, Tim Verbelen, Pieter Simoens, Bart Dhoedt

专题命中隐私与版权：trustworthy(abstract)；分类 cs.LG

Comments ICML 2018 Privacy in Machine Learning and Artificial Intelligence workshop

2506.20481 2025-06-26 cs.LG cs.AI cs.CL cs.CR 56%

Counterfactual Influence as a Distributional Quantity

Matthieu Meeus, Igor Shilov, Georgios Kaissis, Yves-Alexandre de Montjoye

机构 * Imperial College London（帝国理工学院伦敦分校）； Google DeepMind（谷歌DeepMind）

专题命中隐私与版权：分类 cs.CL、cs.AI、cs.LG；trustworthy(comments)

Comments Workshop on The Impact of Memorization on Trustworthy Foundation Models (MemFM) @ ICML 2025

2212.07508 2022-12-16 cs.LG cs.AI cs.CY cs.HC 56%

Tensions Between the Proxies of Human Values in AI

Teresa Datta, Daniel Nissani, Max Cembalest, Akash Khanna, Haley Massa, John P. Dickerson

专题命中隐私与版权：分类 cs.AI、cs.CY、cs.LG；trustworthy(comments)

Comments Contributed Talk, NeurIPS 2022 Workshop on Algorithmic Fairness through the Lens of Causality and Privacy; To be published in 2023 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)

2608.14273 2026-08-17 cs.HC 新提交 50%

Designing Mobile and Wearable Sensor-Fused Conversational Agents for Health and Wellbeing

面向健康与福祉的移动及可穿戴传感器融合对话智能体设计

Hansoo Lee, Pablo Fonseca, Md Haseen Akhtar

专题命中隐私与版权：safety(abstract)

AI总结本教程面向健康领域，旨在教授参与者使用WSDWAS工具，将可穿戴传感器数据与LLM驱动的对话智能体结合，实现从被动监测到可操作福祉对话的转变。

Comments 6 pages, 3 figures. Accepted as a Tutorial at the 28th International Conference on Mobile Human-Computer Interaction (MobileHCI '26)

2603.24735 2026-08-10 cs.HC 版本更新 50%

Examining the Effect of Explanations of AI Privacy Redaction in AI-mediated Interactions

考察AI隐私擦除解释在AI中介互动中的影响

Roshni Kaushik, Maarten Sap, Koichi Onoue

专题命中隐私与版权：trustworthy(abstract)

AI总结研究探讨AI中介互动中解释擦除操作对用户信任的影响，发现解释能提升隐私保护效果，且情境因素和个体差异影响解释效果。

Comments Accepted at AIES 2026

2608.04755 2026-08-06 cs.CR 新提交 50%

"Allow" to Achieve, Over-Privileged Inadvertently: The Unintended Cost of Task-Completion-Driven Pop-up Decisions in Mobile GUI Agents

“允许”达成目标：移动GUI智能体中任务完成驱动的弹窗决策的意外代价

Dongsheng Chen, Yuxuan Li, Guanhua Chen, Jiaxin Zhang, Xiangyu Zhao, Lei Ma, Xin Yao, Xuetao Wei

专题命中隐私与版权：safety(abstract)

AI总结研究移动GUI智能体的权限素养，发现其存在应用信任偏差、任务优先级覆盖等问题，提示干预效果不一，建议将任务执行与权限授权分离。

2506.18266 2026-08-06 cs.CV 版本更新 50%

YouTube-Occ: Learning Indoor 3D Semantic Occupancy Prediction from YouTube Videos

YouTube-Occ：从YouTube视频学习室内3D语义占据预测

Haoming Chen, Lichen Yuan, TianFang Sun, Jingyu Gong, Xin Tan, Zhizhong Zhang, Yanyun Qu, Yuan Xie

机构 * East China Normal University（华东师范大学）

专题命中隐私与版权：alignment(abstract)

AI总结针对室内3D语义占据预测数据稀缺的问题，提出YouTube-Occ框架，通过自动化数据流水线与双对齐预训练策略，在NYUv2等基准上实现性能提升，代码数据将公开。

Comments Accepted by ECCV 2026

2607.26110 2026-07-30 cs.CC cs.SE 新提交 50%

A literature review of recent advances in software design and architecture

软件设计与架构的最新进展文献综述

Malach Obisa Amonga

专题命中隐私与版权：trustworthy(abstract)

AI总结本文综述2024-2025年软件设计与架构研究，采用主题综合法分析五大领域，指出现代架构需覆盖全生命周期，AI辅助等技术可提升软件特性，同时存在实证验证不足等缺口。

2507.18036 2026-07-27 cs.CR cs.CV 版本更新 50%

NWaaS: A Non-Intrusive and Privacy-Preserving Watermarking-as-a-Service System with Adaptive Resource Scheduling

NWaaS：一种具有自适应资源调度的非侵入性和隐私保护水印即服务系统

Haonan An, Qianyao Ren, Guang Hua, Tao Li, Yu Guo, Yanan Ma, Hangcheng Cao, Yuguang Fang

专题命中隐私与版权：trustworthy(abstract)

AI总结研究针对机器学习即服务中保护知识产权的挑战，提出NWaaS框架。核心方法包括$\mathtt{ShadowMark}$算法、协作分区机制和比例差异联合调度算法。贡献是能在多样模态中提供强大所有权验证，保障隐私并提升系统性能。

2607.16280 2026-07-21 cs.CV 新提交 50%

3D FaceShell: Attribute Transfer in 3D Face Avatars as a VLM Defense Mechanism

3D FaceShell：作为视觉语言模型防御机制的3D人脸头像属性转移

Weston Bondurant, Srijan Das, Hieu Le, Stephanie Schuckers

机构 * University of North Carolina at Charlotte（北卡罗来纳大学夏洛特分校）

专题命中隐私与版权：alignment(abstract)

AI总结针对VLM可从3D人脸渲染图像提取敏感属性的隐私挑战，提出3D FaceShell框架，通过可学习高斯壳产生扰动，在保持面部外观的同时引导VLM语义解释，实验证明其能有效提高属性注入和不匹配率。

Comments ECCV 2026

2607.13386 2026-07-16 cs.CV 新提交 50%

FM$^2$: Unified Federated Foundation Models for Heterogeneous Multimodal Medical Imaging

FM$^2$：用于异构多模态医学成像的统一联邦基础模型

Shengchao Chen, Ting Shu

机构 * School of Artificial Intelligence, Shenzhen University（深圳大学人工智能学院）； Australian AI Institute, University of Technology Sydney（悉尼科技大学澳大利亚人工智能研究所）

专题命中隐私与版权：alignment(abstract)

AI总结针对医学成像基础模型构建中隐私与任务统一问题，提出FM$^2$框架，通过从头训练核心主干、结合预训练编码器、配备双混合专家模块及正则化器，并引入字幕增强学习，实现跨模态泛化，优于现有联邦基线。

Comments Accepted by ACM MM 2026 (Main Track): the 34th ACM International Conference on Multimedia

2607.06687 2026-07-09 cs.HC 新提交 50%

Exploring the Interaction of Explanation Styles, Context, and Trust of AI Privacy Redaction in AI-mediated Interactions

探索人工智能隐私编辑的解释风格、上下文和信任在人工智能介导交互中的相互作用

Roshni Kaushik, Maarten Sap, Koichi Onoue

专题命中隐私与版权：trustworthy(abstract)

AI总结研究人工智能介导通信中隐私编辑解释对用户信任的影响，设计场景并开展用户研究，发现提供解释能提升信任，上下文因素有影响，解释偏好因个体差异而异，强调平衡透明度与隐私及适应性解释对设计可信AI系统的重要性。

Comments Originally submitted to UIST, will be resubmitting

2607.03688 2026-07-07 cs.CR 新提交 50%

PathMark: Protecting Intellectual Property of Mixture-of-Expert LLMs via Path Watermarks

PathMark：通过路径水印保护混合专家语言模型的知识产权

Yudong Gao, Qingyue Wang, Yuanyuan Yuan, Ruixuan Huang, Linghan Chen, Zimo Ji, Shuai Wang

专题命中隐私与版权：alignment(abstract)

AI总结针对混合专家（MoE）架构中传统水印方案失效的问题，提出PathMark框架，通过主动控制路由作为隐蔽水印通道，利用多种机制解决脆弱决策边界和路由纠缠问题，实现高验证准确率和强鲁棒性。

Comments 20 pages, accepted by CCS'26

2607.01019 2026-07-02 cs.CR 新提交 50%

Toward a Unified Security and Privacy Framework for AI-Native 6G Networks

面向AI原生6G网络的统一安全与隐私框架

Bidushi Barua, Ahsan Khan, Kangfeng Ye, Panagiotis Papanastasiou, Yifan Liu, Mohit Bidikar, Anthony Moulds, Julie McCann, Poonam Yadav

专题命中隐私与版权：trustworthy(abstract)

AI总结针对AI原生6G网络中安全与隐私的碎片化问题，提出跨层威胁分类与统一框架，分析关键技术威胁并映射对策，为可信6G系统提供基础。

1. 隐私与版权 686 篇

Federated Learning in Vehicular Networks

Distributed learning optimisation of Cox models can leak patient data: Risks and solutions

FlowSense: Monitoring Airflow in Building Ventilation Systems Using Audio Sensing

Enhancing the Security & Privacy of Wearable Brain-Computer Interfaces

Towards the Internet of Behaviors in airports with a fog-to-cloud approach

Evaluating Privacy-Preserving Machine Learning in Critical Infrastructures: A Case Study on Time-Series Classification

CDCGen: Cross-Domain Conditional Generation via Normalizing Flows and Adversarial Training

Model Extraction and Adversarial Attacks on Neural Networks using Switching Power Information

Robust error bounds for quantised and pruned neural networks

SecureBoost: A Lossless Federated Learning Framework

On the Privacy Risks of Model Explanations

Towards Compliant Data Management Systems for Healthcare ML

Asymmetrical Vertical Federated Learning

FedVision: An Online Visual Object Detection Platform Powered by Federated Learning

Privacy Preserving Location Data Publishing: A Machine Learning Approach

Broadband Analog Aggregation for Low-Latency Federated Edge Learning (Extended Version)

Privacy Aware Offloading of Deep Neural Networks

Counterfactual Influence as a Distributional Quantity

Tensions Between the Proxies of Human Values in AI

Designing Mobile and Wearable Sensor-Fused Conversational Agents for Health and Wellbeing

Examining the Effect of Explanations of AI Privacy Redaction in AI-mediated Interactions

"Allow" to Achieve, Over-Privileged Inadvertently: The Unintended Cost of Task-Completion-Driven Pop-up Decisions in Mobile GUI Agents

YouTube-Occ: Learning Indoor 3D Semantic Occupancy Prediction from YouTube Videos

A literature review of recent advances in software design and architecture

NWaaS: A Non-Intrusive and Privacy-Preserving Watermarking-as-a-Service System with Adaptive Resource Scheduling

3D FaceShell: Attribute Transfer in 3D Face Avatars as a VLM Defense Mechanism

FM$^2$: Unified Federated Foundation Models for Heterogeneous Multimodal Medical Imaging

Exploring the Interaction of Explanation Styles, Context, and Trust of AI Privacy Redaction in AI-mediated Interactions

PathMark: Protecting Intellectual Property of Mixture-of-Expert LLMs via Path Watermarks

Toward a Unified Security and Privacy Framework for AI-Native 6G Networks