Author(s): G Karunakar, Huzaifa Tarannum, K Nandana Sri and G Shravani

Abstract: Under sensor assaults, this study explores the cyber-security dilemma of autonomous cars. Secure localization of autonomous cars is a top priority, thus we provide a model-based system that can detect sensor assaults and pinpoint their origins. Introducing sensor redundancy, or the deployment of numerous sensors, each of which offers real-time posture observations of the vehicle, ensures that the vehicle is resistant against cyber-attacks. A set of attack detectors, which includes an extra Kalman filter (EKF) and the cumulative sum (CUSUM) discriminator, is created to identify outliers in every sensor reading. Using EKFs, we can recursively estimate the vehicle's position and orientation. Then, to find any discrepancy between the sensor reading and the predicted pose based on the vehicle's mathematical model, we can use each CUSUM discriminator to analyse the residual produced by its combined EKF. The introduction of an auxiliary detector that combines data from many sensors allows for the monitoring of inconsistencies in the results obtained from these sensors. An isolation strategy based on rules is created to find the source of the aberrant sensor using the data from all the detectors. Using actual car data, we proved that our suggested architecture works.

DOI: 10.33545/26633582.2024.v6.i2b.128

Pages: 91-95 | Views: 892 | Downloads: 454

Download Full Article: Click Here