Recently, the accepted papers from CCS2021’ is published here. I will summarize the related papers in the blog on machine learning. The details from website are here.
Three Lectures:
- Pseudo-Randomness and the Crystal Ball/Cynthia Dwork, Harvard University
- Towards Building a Responsible Data Economy/Dawn Song, University of California, Berkeley
- Are we done yet? Our journey to fight against memory-safety bugs/Taesoo Kim, Georgia Institute of Technology & Samsung Research
Machine Learning and Security 1: Attacks on Robustness
- Black-box Adversarial Attacks on Commercial Speech Platforms with Minimal Information
-
A Hard Label Black-box Adversarial Attack Against Graph Neural Networks
- Robust Adversarial Attacks Against DNN-Based Wireless Communication Systems
- AI-Lancet: Locating Error-inducing Neurons to Optimize Neural Networks
Machine Learning and Security 2: Defenses for ML Robustness
- Learning Security Classifiers with Verified Global Robustness Properties
- On the Robustness of Domain Constraints
- Cert-RNN: Towards Certifying the Robustness of Recurrent Neural Networks
- TSS: Transformation-Specific Smoothing for Robustness Certification
Privacy and Anonymity 1: Inference Attacks
- Honest-but-Curious Nets: Sensitive Attributes of Private Inputs Can Be Secretly Coded into the Classifiers’ Outputs
- Quantifying and Mitigating Privacy Risks of Contrastive Learning
- Membership Inference Attacks Against Recommender Systems
- Membership Leakage in Label-Only Exposures
- When Machine Unlearning Jeopardizes Privacy
1. A Hard Label Black-box Adversarial Attack Against Graph Neural Networks
Author: