DAIRE: Lightweight AI model for real-time CAN attack detection in IoV
Researchers have introduced DAIRE (Detecting Attacks in IoV in REal-time), an efficient machine learning framework designed for the immediate detection and classification of Controller Area Network (CAN) attacks within the Internet of Vehicles (IoV). Utilizing a streamlined artificial neural network (ANN), DAIRE structures each layer with Ni = i x c neurons (where i represents the layer index and c denotes the number of attack classes). Empirical methods are used to set other hyperparameters for real-time functionality. The model focuses on various attacks, including Denial-of-Service, Fuzzy, and Spoofing, and utilizes sparse categorical cross-entropy loss. This framework tackles significant security vulnerabilities in CAN communication, which is prone to cyber threats. The findings are available on arXiv (ID: 2604.20771).
Key facts
- DAIRE stands for Detecting Attacks in IoV in REal-time
- It is a lightweight machine learning framework for real-time detection and classification of CAN attacks
- Built on a lightweight ANN with Ni = i x c neurons per layer
- Ni = number of neurons in ith layer, c = total number of attack classes
- Hyperparameters determined empirically for real-time operation
- Targets attacks: Denial-of-Service, Fuzzy, Spoofing
- Uses sparse categorical cross-entropy loss
- Published on arXiv with ID 2604.20771
Entities
Institutions
- arXiv