Defense Advanced Research Projects AgencyTagged Content List

Physical Security

Relating to the selective release of information and access to facilities

Showing 2 results for Security + Programs RSS
Researchers have demonstrated effective attacks on machine learning (ML) algorithms. These attacks can cause high-confidence misclassifications of input data, even if the attacker lacks detailed knowledge of the ML classifier algorithm and/or training data. Developing effective defenses against such attacks is essential if ML is to be used for defense, security, or health and safety applications.
Electronic system security has become an increasingly critical area of concern for the DoD and more broadly for security of the U.S. as a whole. Current efforts to provide electronic security largely rely on robust software development and integration. Present responses to hardware vulnerability attacks typically consist of developing and deploying patches to the software firewall without identifying or addressing the underlying hardware vulnerability. As a result, while a specific attack or vulnerability instance is defeated, creative programmers can develop new methods to exploit the remaining hardware vulnerability and a continuous cycle of exploitation, patching, and subsequent exploitations ensues.