Vehicle break-ins have reached an epidemic level in recent years in some urban areas. This criminal activity which typically involves breaking vehicle windows and stealing personal properties inside the vehicle only takes a few seconds to complete. However, it can cause a significant property loss and identity theft. In this paper, a parking mode security surveillance system that deters and records the vehicle break-ins is proposed. This system takes advantage of state of the art of internet of things (IoT) and artificial intelligence (AI) technology, making it perform more proactively than traditional shock sensor-based or motion detection-based vehicle security alarm. A low-cost and low-power consumptive RISC-V system which contains basic video/audio signal neural network processing units can serve well in this application.