Evaluation

We have implemented a prototype of our design on USRP2 software radios. We use 400 MHz daughterboards for compatibility with the 402–405 MHz Medical Implant Communication Services (MICS) band used by IMDs. We evaluate our prototype shield against two modern IMDs, namely the Medtronic Virtuoso implantable cardiac defibrillator (ICD) and the Concerto cardiac resynchronization therapy device (CRT). Our evaluation reveals the following:




• When the shield is present, it jams the IMD’s messages, causing even nearby (20 cm away) eavesdroppers to experience a bit error rate of nearly 50%, which is no better than a random guess.








• When the shield jams the IMD’s packets, it can still reliably decode them (the packet loss rate is 0.2%, which is negligible). We conclude that the shield and the IMD share an information channel that is inaccessible to other parties.






• When the shield is absent, the IMD replies to unauthorized commands, even if the adversary is in a non-line-of-sight location more than 14 m away, and uses a commercial device that operates in the MICS band and adheres to the FCC power limit.


• When the shield is present and has the same transmit power as the adversary, the IMD does not respond to unauthorized commands, even when the adversary is only 20 cm away.



• When the shield is absent and an adversary with 100 times the shield’s power transmits unauthorized commands, the IMD responds from distances as large as 27 m.



• When the shield is present the high-powered adversary’s attempts succeed only from distances less than 5 m, and only in line-ofsight locations. The shield always detects high-powered adversarial transmissions and raises an alarm. We conclude that sufficiently high-powered adversaries present an intrinsic limitation to our physical-layer protection mechanism. However, the shield’s presence reduces the adversary’s success range and informs the patient, raising the bar for the adversary’s attempts.