Autonomous vehicles are becoming increasingly popular, but their reliance on computer systems to sense and operate in the physical world has introduced new security risks. Recent studies have shown that using Cache-based Side-Channel Attacks (SCAs) could infer sensitive users' information (e.g., which route the user is taking) highlighting significant vulnerability posed to today's computer systems. In response, we first identify the threat model and victim applications of autonomous driving systems in this work. Next, we explore the suitability of various machine learning-based classifiers trained by information collected from built-in hardware performance counter registers available in modern autonomous vehicle systems. Our experiments conducted on an Intel Xeon, which Waymo autonomous driving vendor uses, demonstrate that J48 achieves 99.5% accuracy with the highest efficiency compared with other investigated models.