A team of researchers from the University of Maryland, Baltimore County (UMBC) have developed a technique for improving the safety of technologically advanced cars vulnerable to malicious network attacks.

To identify the potential of a network attack conducted by a malicious entity who seeks to gain control of an advanced vehicle via intra-vehicular communication networks, the research team created a graph-based anomaly detection technique that demonstrates the relationship between data.

Most vehicles currently rely on controller area networks (CANs) as the vehicle’s built-in intra-vehicular communication network. Although CANs are reportedly simple to use, performing as a broadcasting network, any entity can “read” messages from the car and send conflicting ones in response.

These networks also leave the vehicle vulnerable to malicious entities assuming control over the vehicle, sending new commands that create potentially dangerous scenarios, such as causing engine failure or brake failure.

According to the developers of the graph-based anomaly detection technique, the graphs can be used to perform statistical analysis to identify intruders or potential threats ahead of an attack.

The research appears in the Institution of Electrical and Electronic Engineers (IEEE) publication Transactions on Intelligent Transportation Systems.