Algorithms
Publication
Differentially Private Set Union
Microsoft Research Blog
Newly discovered principle reveals how adversarial training can perform robust deep learning
In machine learning, adversarial examples usually refer to natural inputs plus small, specially crafted perturbations that can fool the model into making mistakes. In recent years, adversarial examples have been repeatedly discovered in deep learning…
Project
DNS Exfiltration Detection
Develop characteristic features of malicious DNS queries and Build anomaly detection models
Event
Microsoft at STOC 2020
Microsoft is proud to be a sponsor of the 52nd Annual ACM Symposium on Theory of Computing (STOC). See detailed information on Microsoft’s contributions to the program.
