The potential for large language models (LLMs) to hide messages within plain text (steganography) poses a challenge to detection and thwarting of unaligned AI agents, and undermines faithfulness of LLMs reasoning. We explore the steganographic capabilities of LLMs fine-tuned via reinforcement learning (RL) to: (1) develop covert encoding schemes, (2) engage in steganography when prompted, and (3) utilize steganography in realistic scenarios where hidden reasoning is likely, but not prompted. In these scenarios, we detect the intention of LLMs to hide their reasoning as well as their steganography performance. Our findings in the fine-tuning experiments as well as in behavioral non fine-tuning evaluations reveal that while current models exhibit rudimentary steganographic abilities in terms of security and capacity, explicit algorithmic guidance markedly enhances their capacity for information concealment.
MCTSteg: A Monte Carlo Tree Search-based Reinforcement Learning Framework for Universal Non-additive Steganography
College of Computer Science and Software Engineering, Shenzhen University（深圳大学计算机与软件学院）；College of Information Engineering, Shenzhen University（深圳大学信息工程学院）；Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ)（广东省人工智能与数字经济实验室（深圳））；Shenzhen Institute of Artificial Intelligence and Robotics for Society（深圳市人工智能与机器人研究院）；Alibaba Group（阿里巴巴集团）
Large language models can learn and generalize steganographic chain-of-thought under process supervision
Mentorship for Alignment Research Students (MARS)（对齐研究 mentorship 项目）；University College London（伦敦大学学院）；Queen Mary University of London（伦敦女王学院）；ML Alignment & Theory Scholars (MATS)（对齐与理论学者）；Meridian Impact, Cambridge（剑桥 Meridian Impact）；Universidad Carlos III de Madrid（马德里卡洛斯三世大学）；Geodesic Research and University of Cambridge（Geodesic Research 和剑桥大学）
A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring
University of Cambridge（剑桥大学）；Massachusetts Institute of Technology（麻省理工学院）；UK AI Safety Institute（英国人工智能安全研究所）；University of Oxford（牛津大学）；Mila, University of Montreal（蒙特利尔大学米尔人工智能实验室）
Hide and Seek in Embedding Space: Geometry-based Steganography and Detection in Large Language Models
UCL Centre for Artificial Intelligence, University College London, UK（伦敦大学学院人工智能中心，大学学院伦敦）；School of Computing and Communications, Lancaster University, UK（兰卡斯特大学计算机与通讯学院）；Department of Informatics, University of Sussex, UK（苏塞克斯大学信息学院）；Centre for Psychedelic Research and Centre for Complexity Science, Imperial College London, UK（伦敦帝国学院迷幻研究与复杂科学中心）；Centre for Eudaimonia and Human Flourishing, University of Oxford, UK（牛津大学幸福与人类繁荣中心）
Exploring the Potential of Large Language Models for Improving Digital Forensic Investigation Efficiency
Security Research Group, School of Computer Science , addressline= University College Dublin , city= Belfield , state= Dublin 4 , country= Ireland；organization= School of Criminal Justice , addressline= University of Lausanne , state= Lausanne , country= Switzerland；organization= Present affiliation: Institute of Computer Science , addressline= University of Augsburg , state= Augsburg , country= Germany
Applying Large Language Models to Power Systems: Potential Security Threats
The Hong Kong Polytechnic University（香港理工大学）；Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））；The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；The University of Sydney（悉尼大学）；Zhejiang University（浙江大学）；Nanyang Technological University（南洋理工大学）
Unlocking the Potential of Model Merging for Low-Resource Languages
Wangxuan Institute of Computer Technology, Peking University（北京大学王选计算机研究所）；State Key Laboratory of General Artificial Intelligence, Peking University（北京大学通用人工智能国家重点实验室）；Center for Data Science, Peking University（北京大学数据科学中心）；College of Engineering, Peking University（北京大学工学院）
Unveiling the Potential of Sentiment: Can Large Language Models Predict Chinese Stock Price Movements?
International Digital Economy Academy (IDEA)（国际数字经济学院（IDEA））；The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；Data Science and Analytics Thrust, The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州）数据科学与分析学域）；Financial Technology Thrust, The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州）金融科技学域）
Institute of Computational Life Sciences, Zurich University of Applied Sciences（计算生命科学学院，应用科学大学）；Institute of History and Philosophy of Science and Technology (IHPST), Université Paris 1 - Panthéon-Sorbonne（科学与技术历史与哲学研究所，巴黎第一大学）
ChatGPT for Shaping the Future of Dentistry: The Potential of Multi-Modal Large Language Model
State Key Laboratory of Oral Diseases & National Clinical Research Center for Oral Diseases & Department of Oral and Maxillofacial Surgery, West China Hospital of Stomatology, Sichuan University（四川大学华西口腔医院口腔疾病研究国家重点实验室、国家口腔疾病临床医学研究中心、口腔颌面外科）；Department of Civil, Environmental & Construction Engineering, University of Central Florida（中佛罗里达大学土木、环境与建筑工程系）；College of Transportation Engineering, University of Central Florida（中佛罗里达大学交通工程学院）；School of Transportation and Logistics, Southwest Jiaotong University（西南交通大学交通运输与物流学院）；C2SMART Center, Tandon School of Engineering, New York University（纽约大学坦登工程学院C2SMART中心）；State Key Laboratory of Oral Diseases & National Clinical Research Center for Oral Diseases & Eastern Clinic, West China Hospital of Stomatology, Sichuan University（四川大学华西口腔医院口腔疾病研究国家重点实验室、国家口腔疾病临床医学研究中心、东院区）
Leveraging the Potential of Prompt Engineering for Hate Speech Detection in Low-Resource Languages
Bangladesh Institute of Governance and Management (BIGM)（孟加拉国治理与管理研究所）；Department of Cyber Security Engineering, George Mason University（网络安全工程系，乔治·梅森大学）
Generative Language Models Potential for Requirement Engineering Applications: Insights into Current Strengths and Limitations
Rhineland-Palatinate Technical University of Kaiserslautern-Landau（莱茵兰-普法尔茨州凯撒斯劳滕-兰道技术大学）；German Research Center for Artificial Intelligence GmbH（德国人工智能研究中心有限公司）
