Description

本日紹介した論文の一覧

• Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity
  Analysis Between Alignment and Fine-tuning Datasets
  http://arxiv.org/abs/2506.05346v1

• Big Bird: Privacy Budget Management for W3C's Privacy-Preserving
  Attribution API
  http://arxiv.org/abs/2506.05290v1

• SECNEURON: Reliable and Flexible Abuse Control in Local LLMs via Hybrid
  Neuron Encryption
  http://arxiv.org/abs/2506.05242v1

• OpenCCA: An Open Framework to Enable Arm CCA Research
  http://arxiv.org/abs/2506.05129v1

• Membership Inference Attacks on Sequence Models
  http://arxiv.org/abs/2506.05126v1

• Privacy Amplification Through Synthetic Data: Insights from Linear
  Regression
  http://arxiv.org/abs/2506.05101v1

• EMBER2024 -- A Benchmark Dataset for Holistic Evaluation of Malware
  Classifiers
  http://arxiv.org/abs/2506.05074v1

• Identifying and Understanding Cross-Class Features in Adversarial
  Training
  http://arxiv.org/abs/2506.05032v1

• Tech-ASan: Two-stage check for Address Sanitizer
  http://arxiv.org/abs/2506.05022v1

• Attack Effect Model based Malicious Behavior Detection
  http://arxiv.org/abs/2506.05001v1

• Evaluating the Impact of Privacy-Preserving Federated Learning on CAN
  Intrusion Detection
  http://arxiv.org/abs/2506.04978v1

• Hiding in Plain Sight: Query Obfuscation via Random Multilingual
  Searches
  http://arxiv.org/abs/2506.04963v1

• PoCGen: Generating Proof-of-Concept Exploits for Vulnerabilities in Npm
  Packages
  http://arxiv.org/abs/2506.04962v1

• When Thinking LLMs Lie: Unveiling the Strategic Deception in
  Representations of Reasoning Models
  http://arxiv.org/abs/2506.04909v1

• A Private Smart Wallet with Probabilistic Compliance
  http://arxiv.org/abs/2506.04853v1

• On Automating Security Policies with Contemporary LLMs
  http://arxiv.org/abs/2506.04838v1

• MULTISS: un protocole de stockage confidentiel {à} long terme sur
  plusieurs r{é}seaux QKD
  http://arxiv.org/abs/2506.04800v1

• Urania: Differentially Private Insights into AI Use
  http://arxiv.org/abs/2506.04681v1

• Authenticated Private Set Intersection: A Merkle Tree-Based Approach for
  Enhancing Data Integrity
  http://arxiv.org/abs/2506.04647v1

• Incentivizing Collaborative Breach Detection
  http://arxiv.org/abs/2506.04634v1

• BESA: Boosting Encoder Stealing Attack with Perturbation Recovery
  http://arxiv.org/abs/2506.04556v1

なお、ポッドキャスト内で紹介する内容は、各論文の概要を日本語で解説したもので、論文概要の著作権は論文著者に帰属します。