作者
Dawn Song
AI / Security
BEEAR: Embedding-based Adversarial Removal of Safety Backdoors in Instruction-tuned Language Models
Evolving AI Collectives to Enhance Human Diversity and Enable Self-Regulation
Comments ICML 2024
Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression
Comments Accepted to ICML'24
AI Risk Management Should Incorporate Both Safety and Security
TrojFM: Resource-efficient Backdoor Attacks against Very Large Foundation Models
Managing extreme AI risks amid rapid progress
Comments Published in Science: https://www.science.org/doi/10.1126/science.adn0117
Effective and Efficient Federated Tree Learning on Hybrid Data
KnowHalu: Hallucination Detection via Multi-Form Knowledge Based Factual Checking
Berkeley Open Extended Reality Recordings 2023 (BOXRR-23): 4.7 Million Motion Capture Recordings from 105,852 Extended Reality Device Users
Comments Learn more at https://rdi.berkeley.edu/metaverse/boxrr-23
Journal ref IEEE Transactions on Visualization and Computer Graphics, pages 1-8, March 2024. IEEE VR 2024, Orlando, FL March 16-21, 2024. Best Paper Honorable Mention
Benchmarking Zero-Shot Robustness of Multimodal Foundation Models: A Pilot Study
On the Societal Impact of Open Foundation Models
DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models
Comments NeurIPS 2023 Outstanding Paper (Datasets and Benchmarks Track)
The Blockchain Imitation Game
GRATH: Gradual Self-Truthifying for Large Language Models
DiffAttack: Evasion Attacks Against Diffusion-Based Adversarial Purification
Comments Accepted to NeurIPS 2023
Specular: Towards Secure, Trust-minimized Optimistic Blockchain Execution
UniFed: All-In-One Federated Learning Platform to Unify Open-Source Frameworks
Comments Code: https://github.com/AI-secure/FLBenchmark-toolkit Website: https://unifedbenchmark.github.io/
Identifying and Mitigating the Security Risks of Generative AI
Journal ref Foundations and Trends in Privacy and Security 6 (2023) 1-52
Exploring the Privacy Risks of Adversarial VR Game Design
Comments Learn more at https://rdi.berkeley.edu/metaverse/metadata
Journal ref 23rd Privacy Enhancing Technologies Symposium (2023) 238-256
TextGuard: Provable Defense against Backdoor Attacks on Text Classification
Comments Accepted by NDSS Symposium 2024
Deep Motion Masking for Secure, Usable, and Scalable Real-Time Anonymization of Virtual Reality Motion Data
Truth in Motion: The Unprecedented Risks and Opportunities of Extended Reality Motion Data
Journal ref IEEE Security & Privacy (2024)
SoK: Data Privacy in Virtual Reality
Journal ref 24th Privacy Enhancing Technologies Symposium (2024) 21-40
Unique Identification of 50,000+ Virtual Reality Users from Head & Hand Motion Data
Journal ref 32nd USENIX Security Symposium (2023) 895-910
Going Incognito in the Metaverse: Achieving Theoretically Optimal Privacy-Usability Tradeoffs in VR
Comments Learn more at https://rdi.berkeley.edu/metaverse/metaguard/
Journal ref 36th Annual ACM Symposium on User Interface Software and Technology (2023)
Multi-Factor Credential Hashing for Asymmetric Brute-Force Attack Resistance
Journal ref 8th IEEE European Symposium on Security and Privacy (2023) 56-72
Decentralizing Custodial Wallets with MFKDF
Journal ref 5th IEEE International Conference on Blockchain and Cryptocurrency (2023) 1-9
Multi-Factor Key Derivation Function (MFKDF) for Fast, Flexible, Secure, & Practical Key Management
Comments To appear in USENIX Security '23
Journal ref 32nd USENIX Security Symposium (2023) 2097-2114
SoK: Privacy-Preserving Data Synthesis
Comments Accepted at IEEE S&P (Oakland) 2024