Back to search

IKTPLUSS-IKT og digital innovasjon

AI-DRIVEN TESTING OF FALSE DATA INJECTION ATTACKS AGAINST TRANSPORT INFRASTRUCTURES

Alternative title: AI-DRIVEN TESTING OF FALSE DATA INJECTION ATTACKS AGAINST TRANSPORT INFRASTRUCTURES

Awarded: NOK 10.0 mill.

The TSAR project developed a novel approach for the detection of one type of false data injection attack in the maritime, namely, an intentional AIS shutdown in the open sea. Detecting this type of attack can be useful for identifying illegal activities at sea, such as illegal fishing or transshipment of illicit goods. The developed approach is based on self-supervised deep learning techniques and transformer models. It uses historical data to train the model that predicts if an AIS message should be received in the upcoming minute. The model reports detected anomalies by comparing the prediction with observation. The approach can process AIS messages in real-time, in particular, more than 500 Million AIS messages per month, corresponding to the trajectories of more than 60 000 ships. The method is evaluated on 1-year of real-world data coming from four Norwegian surveillance satellites.

The project developed a novel approach for detecting intentional AIS shutdown in open sea maritime surveillance using self-supervised deep learning, which can be useful for detecting illegal activities such as illegal fishing or smuggling. The approach has been validated in a use case containing more than 500 million AIS messages per month, corresponding to the trajectories of more than 60 000 ships, collected during one year of real-world data collection from four Norwegian surveillance satellites.

In transport infrastructures, vessel traffic services, air-traffic management and connected cars all rely on unauthenticated and unencrypted messages transfer that renders these services vulnerable to cyberattacks. Typical attacks such as False Data Injection Attacks (FDIA) are difficult to detect as they alter the semantics of the data (e.g., by adding/removing/multiplying elements on a real-time control equipment), while preserving the syntactical correctness of the messages. Identifying these attacks and classifying them as serious threats or unintentional false data has become a major challenge of traffic monitoring authorities. The TSAR project aims at demonstrating that recent advances in Artificial Intelligence (AI) can be leveraged in the automatic detection of FDIA in transport infrastructures. By combining realistic threat data generation based on constraint-based software testing techniques and automatic detection with deep reinforcement learning, TSAR will propose a new technology for automatic FDIA generation and detection. This technology will be empirically evaluated with end-users from the maritime domain and with open and accessible data in two other domains, namely air traffic control and connected cars. By leveraging automatic detection of FDIA in traffic management systems, TSAR will also prepare the ground for the upcoming revolution in traffic management which concerns, self-driving vessels, self-driving aircrafts and self-driving cars.

Funding scheme:

IKTPLUSS-IKT og digital innovasjon