TY - JOUR
T1 - Electromagnetic Fingerprinting of Memory Heartbeats
T2 - System and Applications
AU - SHEN, Cheng
AU - HUANG, Jun
AU - SUN, Guangyu
AU - CHEN, Jingshu
PY - 2022/9
Y1 - 2022/9
N2 - This paper presents MemScope, a system that fingerprints devices via electromagnetic sensing of their memory heartbeats, i.e., the clock signal that synchronizes memory and memory controller. MemScope leverages the enhanced resolution and security of memory heartbeat fingerprint, which has enriched spectral features thanks to the spread spectrum generation of memory clock, and cannot be concealed as long as the device accesses its memory during computing. MemScope employs signal processing algorithms that allow it to hear the memory heartbeats of devices from a distance, in the presence of noise, and in crowded environments where multiple devices coexist. It then fingerprints memory heartbeats using machine learning tools. Measurements on a set of 65 devices over a month validate the robustness of fingerprint against time variation, and show a high precision and recall. We then use the neural network to build a detector to defend against possible replay attacks. Finally, we further demonstrate the effectiveness of MemScope in two application scenarios, (i) detecting wireless identity spoofing and (ii) identifying and localizing unauthorized hidden cameras.
AB - This paper presents MemScope, a system that fingerprints devices via electromagnetic sensing of their memory heartbeats, i.e., the clock signal that synchronizes memory and memory controller. MemScope leverages the enhanced resolution and security of memory heartbeat fingerprint, which has enriched spectral features thanks to the spread spectrum generation of memory clock, and cannot be concealed as long as the device accesses its memory during computing. MemScope employs signal processing algorithms that allow it to hear the memory heartbeats of devices from a distance, in the presence of noise, and in crowded environments where multiple devices coexist. It then fingerprints memory heartbeats using machine learning tools. Measurements on a set of 65 devices over a month validate the robustness of fingerprint against time variation, and show a high precision and recall. We then use the neural network to build a detector to defend against possible replay attacks. Finally, we further demonstrate the effectiveness of MemScope in two application scenarios, (i) detecting wireless identity spoofing and (ii) identifying and localizing unauthorized hidden cameras.
KW - device fingerprinting
KW - electromagnetic radiation
KW - side-channel
UR - http://www.scopus.com/inward/record.url?scp=85139165972&partnerID=8YFLogxK
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85139165972&origin=recordpage
U2 - 10.1145/3550295
DO - 10.1145/3550295
M3 - RGC 21 - Publication in refereed journal
SN - 2474-9567
VL - 6
JO - Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
JF - Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
IS - 3
M1 - 138
ER -