Guaranteeing AI Robustness against Deception (GARD)
Abstract
The Guaranteeing AI Robustness against Deception (GARD) program, expanding on technical challenges encountered in the Lifelong Learning Machines program, will develop techniques to make artificial intelligence (AI) and machine learning (ML) algorithms and systems more robust in the presence of deceptive data and adversarial attack. Concurrent with the recent explosion of interest in ML, deception attacks that manipulate a ML system into an erroneous response have also emerged. While such deception attacks against ML have become sophisticated and varied, the development of defensive capabilities for ML systems has not been maintained. The GARD program will address the growing need for defensive ML capabilities by developing techniques to establish robustness properties of ML systems, and to defend against possible attacks. The techniques developed under RAIAD will be essential if the DoD is to rely on ML systems in contested environments.
Document Details
- Document Type
- Accomplishment
- Publication Date
- Oct 01, 2020
- Source ID
- 4b5372df5f506c4fe02fd9e1a3507e74