Xiao Zhang and my paper on Cost-Sensitive Robustness against Adversarial Examples has been accepted to ICLR 2019.
Several recent works have developed methods for training classifiers that are certifiably robust against norm-bounded adversarial perturbations. However, these methods assume that all the adversarial transformations provide equal value for adversaries, which is seldom the case in real-world applications. We advocate for cost-sensitive robustness as the criteria for measuring the classifier’s performance for specific tasks. We encode the potential harm of different adversarial transformations in a cost matrix, and propose a general objective function to adapt the robust training method of Wong & Kolter (2018) to optimize for cost-sensitive robustness. Our experiments on simple MNIST and CIFAR10 models and a variety of cost matrices show that the proposed approach can produce models with substantially reduced cost-sensitive robust error, while maintaining classification accuracy.
A Pragmatic Introduction to Secure Multi-Party Computation, co-authored with Vladimir Kolesnikov and Mike Rosulek, is now published by Now Publishers in their Foundations and Trends in Privacy and Security series.
You can download the book for free (we retain the copyright and are allowed to post an open version) from securecomputation.org, or buy an PDF version from the published for $260 (there is also a printed $99 version).
Bargav Jayaraman presented our work on privacy-preserving machine learning at the 32nd Conference on Neural Information Processing Systems (NeurIPS 2018) in Montreal.
Distributed learning (sometimes known as federated learning) allows a group of independent data owners to collaboratively learn a model over their data sets without exposing their private data. Our approach combines differential privacy with secure multi-party computation to both protect the data during training and produce a model that provides privacy against inference attacks.
We explore two popular methods of differential privacy, output perturbation and gradient perturbation, and advance the state-of-the-art for both methods in the distributed learning setting. In our output perturbation method, the parties combine local models within a secure computation and then add therequired differential privacy noise before revealing the model. In our gradient perturbation method, the data owners collaboratively train a global model via aniterative learning algorithm. At each iteration, the parties aggregate their local gradients within a secure computation, adding sufficient noise to ensure privacy before the gradient updates are revealed. For both methods, we show that the noise can be reduced in the multi-party setting by adding the noise inside the securecomputation after aggregation, asymptotically improving upon the best previous results. Experiments on real world data sets demonstrate that our methods providesubstantial utility gains for typical privacy requirements.
Bargav Jayaraman, Lingxiao Wang, David Evans and Quanquan Gu. Distributed Learning without Distress: Privacy-Preserving Empirical Risk Minimization. 32nd Conference on Neural Information Processing Systems (NeurIPS). Montreal, Canada. December 2018. (PDF, 19 pages, including supplemental materials)
I gave the Booz Allen Hamilton Distinguished Colloquium at the University of Maryland on Can Machine Learning Ever Be Trustworthy?.
The National Science Foundation announced the Center for Trustworthy Machine Learning today, a new five-year SaTC Frontier Center “to develop a rigorous understanding of the security risks of the use of machine learning and to devise the tools, metrics and methods to manage and mitigate security vulnerabilities.”
The Center is lead by Patrick McDaniel at Penn State University, and in addition to our group, includes Dan Boneh and Percy Liang (Stanford University), Kamalika Chaudhuri (University of California San Diego), Somesh Jha (University of Wisconsin) and Dawn Song (University of California Berkeley).