Video-analysis inference automated ECG (VID-ECG): improving video-based heart rate detection and exposing security risks of ECG-based biometric authentication

Many recent biometric authentication methods using heart signals in the form of ECG and its components have been proposed to be used as a unique security key for body area networks (BANs) to authenticate individuals and protect privacy and network security. In this thesis we show how compo- nents of information on cardiac activity, heart rate and beat-to-beat heart pulse information can be extracted easily using our video-based non-contact method and expose the vulnerability of such biometric security protocols. We propose a novel method called Video-analysis Inference Automated ECG (VID-ECG) for pulse extraction by facial video processing. Our al- gorithm combines facial region tracking, motion stabilization, filtering and heart beat information extraction methods to allow automated extraction of each pulse from subject facial videos. VID-ECG results show a high level of accuracy and, unlike related methods in this area, VID-ECG does automatic extraction without knowledge of any frequency range. It is also able to han- dle natural motion in subjects. We applied VID-ECG on a wide range of subjects with varied skin tones, and found accuracy to be high, with more than 0.9 cross-correlation with ground truth and error less than 0.085% of average heart rate for each sample. Results have also been compared with a previously proposed video based method for heart rate extraction, and ac- curacy and beat-to-beat correspondence have been shown to be significantly improved, mainly due to the more realistic filtering used and improved mo- tion handling features of VID-ECG. As we are able to obtain many components of cardiac activity such as average heart rate information and close to real-time beat-to-beat informa- tion, we discuss the implication of our results and how VID-ECG exposes the vulnerability of ECG/cardiac data based biometric authentication meth- ods to remote attack using easily obtainable video data from omnipresent commodity cameras around us today in public and private spaces.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2016

Author dc:creator

• Adhikari, Anku

Contributors dc:contributor

• Hu, Yih-Chun

• security
• biometric
• video processing
• Video-analysis Inference Electrocardiogram (VID-ECG)
• image processing
• signal processing
• Eulerian

Statement dc:rights

• Copyright 2015 Anku Adhikari

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/89206

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/89206