Semantic attacks on wireless medical devices

Security of medical embedded systems is of vital importance. Wireless medical devices used in wireless health applications employ large number of sensors and are in particular susceptible to security attacks. They are often not physically secured and are usually used in hostile environments. We have developed theoretical and statistical framework for creating semantic attacks where data is altered in such a way that the consequences include incorrect medical diagnosis and treatment. Our approach maps a semantic attack to an instance of optimization problem where medical damage is maximized under constraints of the probability of detection and root cause tracing. We use a popular medical shoe to demonstrate that low energy and low cost of embedded medical devices increases the probability of successful attacks. We have proposed two types of semantic attacks, respectively pressure-based attack, and time-based attack under two scenarios, a shoe with 99 pressure sensors and a shoe with 20 pressure sensors. We test the effects of the attacks and compare them. Our results indicate that it is surprisingly easy to attack several essential medical metrics and to alter corresponding medical diagnosis.