A Cloud Collaborative-based Intrusion Detection and Prevention System for IVN
contributation:
- This paper consolidates HCRL’s Car Hacking Dataset (CHD), the Survival Analysis Dataset (SAD) [16], and data collected from real invehicle/simulation environments. This integration results in the development of an In-Vehicle Attack Dataset (IVAD).
- To address the limitations of in-vehicle IDSs due to the restricted computational resources within vehicles, this paper proposes the CC-IDPS. It conducts model training tasks on cloud servers, and subsequently distributes the model to vehicles for detection. The detection model of this system primarily employs BERT for classifying the traffic under inspection. Subsequently, based on the classification results, vehicles undertake appropriate actions to defend against the malicious traffic.
- This study enhances the BERT model for IVN intrusion detection by adapting it to process serialized IVN data, effectively transforming features like message IDs and payload contents into a suitable input format.
- The CC-IDPS involves an Encode2ID algorithm to encode malicious traffic, generating a unique ID for each. As a result of the Encode2ID algorithm, the generated strings occupy less space compared to before, thereby saving storage space and reducing non-repetitive comparison time.
.