TY - JOUR
T1 - "Use It-No Need to Shake It!" Accurate Implicit Authentication for Everyday Objects with Smart Sensing
AU - Wu, Chuxiong
AU - Li, Xiaopeng
AU - Zuo, Fei
AU - Luo, Lannan
AU - Du, Xiaojiang
AU - Di, Jia
AU - Zeng, Qiang
N1 - Publisher Copyright:
© 2022 ACM.
PY - 2022/9/7
Y1 - 2022/9/7
N2 - Implicit authentication for traditional objects, such as doors and dumbbells, has rich applications but is rarely studied. An ongoing trend is that traditional objects are retrofitted to smart environments; for instance, a contact sensor is attached to a door to detect door opening (but cannot tell "who is opening the door"). We present the first accurate implicit-authentication system for retrofitted everyday objects, named MoMatch. It makes an authentication decision based on a single natural object use, unlike prior work that requires shaking objects. MoMatch is built on the observation that an object has a motion typically because a human hand moves it; thus, the object's motion and the legitimate user's hand movement should correlate. The main challenge is, given the small amount of data collected during one object use, how to measure the correlation accurately. We convert the correlation measurement problem into an image comparison problem and resolve it using neural networks successfully. MoMatch does not need to profile the user's biometric information and is resilient to mimicry attacks.
AB - Implicit authentication for traditional objects, such as doors and dumbbells, has rich applications but is rarely studied. An ongoing trend is that traditional objects are retrofitted to smart environments; for instance, a contact sensor is attached to a door to detect door opening (but cannot tell "who is opening the door"). We present the first accurate implicit-authentication system for retrofitted everyday objects, named MoMatch. It makes an authentication decision based on a single natural object use, unlike prior work that requires shaking objects. MoMatch is built on the observation that an object has a motion typically because a human hand moves it; thus, the object's motion and the legitimate user's hand movement should correlate. The main challenge is, given the small amount of data collected during one object use, how to measure the correlation accurately. We convert the correlation measurement problem into an image comparison problem and resolve it using neural networks successfully. MoMatch does not need to profile the user's biometric information and is resilient to mimicry attacks.
KW - Deep Learning
KW - Everyday Objects
KW - Implicit Authentication
KW - Smart Sensing
UR - http://www.scopus.com/inward/record.url?scp=85139267234&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85139267234&partnerID=8YFLogxK
U2 - 10.1145/3550322
DO - 10.1145/3550322
M3 - Article
AN - SCOPUS:85139267234
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 - 146
ER -