CHARLES RIVER ANALYTICS, INC. — Department of Defense SBIR Phase I: The effective functioning of embedded networked devices is critical to the US's natio

Description

The effective functioning of embedded networked devices is critical to the US's national security. A successful malicious attack on these systems could cause widespread damage and possibly deaths. Unfortunately, most existing embedded systems are largely unprotected and vulnerable to attack. A method is needed to detect novel, unforeseen vulnerabilities in an embedded device and patch them without taking the device offline. Charles River Analytics proposes Self-Modeling and Self-Healing Embedded Devices (SMASHED), which use a probabilistic model of the ordinary functioning of the device to detect deviations from expected behavior, as well as potentially damaging deviations from normal input patterns. Unlike traditional statistical anomaly detection approaches, SMASHED judges abnormal behavior within the context of its device model, potentially leading to fewer false positives. SMASHED uses probabilistic programming as a framework for reasoning about the probabilistic model for implementation in hardware or firmware.