D
Dawn Song
Researcher at University of California, Berkeley
Publications - 504
Citations - 75245
Dawn Song is an academic researcher from University of California, Berkeley. The author has contributed to research in topics: Computer science & Deep learning. The author has an hindex of 117, co-authored 460 publications receiving 61572 citations. Previous affiliations of Dawn Song include FireEye, Inc. & University of California.
Papers
More filters
Proceedings ArticleDOI
The Honey Badger of BFT Protocols
TL;DR: HoneyBadgerBFT is presented, the first practical asynchronous BFT protocol, which guarantees liveness without making any timing assumptions, and is based on a novel atomic broadcast protocol that achieves optimal asymptotic efficiency.
Proceedings ArticleDOI
SoK: Eternal War in Memory
TL;DR: The current knowledge about various protection techniques are systematized by setting up a general model for memory corruption attacks, and what policies can stop which attacks are shown, to analyze the reasons why protection mechanisms implementing stricter polices are not deployed.
Proceedings Article
Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty
TL;DR: This work finds that self-supervision can benefit robustness in a variety of ways, including robustness to adversarial examples, label corruption, and common input corruptions, and greatly benefits out-of-distribution detection on difficult, near-dist distribution outliers.
Posted Content
Delving into Transferable Adversarial Examples and Black-box Attacks
TL;DR: In this paper, Xu et al. proposed novel ensemble-based approaches to generate transferable adversarial examples, and observed a large proportion of targeted adversarial instances that are able to transfer with their target labels for the first time.
Proceedings Article
The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks
TL;DR: This paper describes a testing methodology for quantitatively assessing the risk that rare or unique training-data sequences are unintentionally memorized by generative sequence models---a common type of machine-learning model, and describes new, efficient procedures that can extract unique, secret sequences, such as credit card numbers.