Towards Privacy-Preserving Driver’s Drowsiness and Distraction Detection: A Differential Privacy Approach

The ubiquitous need for detecting driver’s drowsiness and distraction by using a camera mounted inside the car directed to driver’s face is driving significant concern about protecting driver’s identity from being discovered or exploited by hackers. In this paper, we present a novel technique called block Laplacian Obfuscation Mechanism (bLOM), to privatize the camera data stream by using differential privacy techniques introduced in database domain. We introduce a metric to measure privacy and utility for driver’s drowsiness and distraction algorithms. Our experimental results show that in 87% of test cases, bLOM is able to keep the identity of the driver private while still be able to extract facial features needed for driver’s drowsiness and distraction detection.

[1]  Anand D. Sarwate,et al.  Signal Processing and Machine Learning with Differential Privacy: Algorithms and Challenges for Continuous Data , 2013, IEEE Signal Processing Magazine.

[2]  M. Amaç Güvensan,et al.  Driver Behavior Analysis for Safe Driving: A Survey , 2015, IEEE Transactions on Intelligent Transportation Systems.

[3]  Charles Elkan,et al.  Differential Privacy and Machine Learning: a Survey and Review , 2014, ArXiv.

[4]  Nicolae Tapus,et al.  Input Validation for the Laplace Differential Privacy Mechanism , 2015, 2015 20th International Conference on Control Systems and Computer Science.

[5]  Aaron Roth,et al.  The Algorithmic Foundations of Differential Privacy , 2014, Found. Trends Theor. Comput. Sci..

[6]  Stefan Katzenbeisser,et al.  Privacy-Preserving Face Recognition , 2009, Privacy Enhancing Technologies.

[7]  Yves Deswarte,et al.  Survey on security threats and protection mechanisms in embedded automotive networks , 2013, 2013 43rd Annual IEEE/IFIP Conference on Dependable Systems and Networks Workshop (DSN-W).

[8]  Pramod Viswanath,et al.  Optimal Noise Adding Mechanisms for Approximate Differential Privacy , 2016, IEEE Transactions on Information Theory.

[9]  Vikas Maheshkar,et al.  FEATURE IMAGE GENERATION USING LOW , MID AND HIGH FREQUENCY REGIONS FOR FACE RECOGNITION , 2012 .

[10]  S. Langton,et al.  The influence of head contour and nose angle on the perception of eye-gaze direction , 2004, Perception & psychophysics.

[11]  H. Poor,et al.  Utility-Privacy Tradeoff in Databases : An Information-theoretic Approach , 2013 .

[12]  Ufuk Topcu,et al.  Differentially private distributed protocol for electric vehicle charging , 2014, 2014 52nd Annual Allerton Conference on Communication, Control, and Computing (Allerton).

[13]  Pramod Viswanath,et al.  The optimal mechanism in differential privacy , 2012, 2014 IEEE International Symposium on Information Theory.

[14]  David Masip,et al.  Preferred Spatial Frequencies for Human Face Processing Are Associated with Optimal Class Discrimination in the Machine , 2008, PloS one.

[15]  Anand D. Sarwate,et al.  Differentially Private Empirical Risk Minimization , 2009, J. Mach. Learn. Res..

[16]  Jun Sakuma,et al.  Differentially Private Empirical Risk Minimization with Input Perturbation , 2017, DS.

[17]  Matti Valovirta,et al.  Experimental Security Analysis of a Modern Automobile , 2011 .

[18]  Ling Huang,et al.  Learning in a Large Function Space: Privacy-Preserving Mechanisms for SVM Learning , 2009, J. Priv. Confidentiality.

[19]  Nader M. Rabadi,et al.  Privacy Protection Among Drivers in Vehicle-to-Vehicle Communication Networks , 2007, 2007 4th IEEE Consumer Communications and Networking Conference.

[20]  George J. Pappas,et al.  Differentially Private Filtering , 2012, IEEE Transactions on Automatic Control.