The invention belongs to the field of intrusion detection algorithms in the Internet of Vehicles, and particularly relates to an Internet-of-Vehicles intrusion detection system based on a hidden Markov model. The system is used for intrusion detection of a False alert attack, a Sybil attack, a Black hole attack and a DoS attack in the Internet of Vehicles. The system mainly comprises a pre-detection module, a DNN-based detection center module, an updating module used for recording a vehicle state and generating a hidden Markov model, and a response center module used for generating a responsesignal, and under the normal operation state of the Internet of Vehicles, the modules jointly guarantee efficient operation of the Internet of Vehicles. Under the attack detection state, the components supplement each other to jointly complete an attack monitoring defense process. Compared with the DNN-based IDS, the Internet-of-Vehicles intrusion detection system has the advantages that the detection precision, overhead and detection time are improved, the detection precision is higher, the average detection time is far shorter than that of an IDS based on DNN, meanwhile, the expenditure is lower, compared with an IDS using state switching, the expenditure of a pre-detection mechanism used in the method is lower, and more computing resources are saved.