Page 826 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

This deep learning-based approach offers a clear advantage over traditional signature-based or
rule-based IDS methods. Traditional IDSs are typically designed for relatively static networks and
rely on known attack patterns, making them ineffective against zero-day attacks. In contrast, our
approach is dynamic, adaptive, and better suited for the mobile, real-time nature of vehicular
communication.
We adopt a dual-dataset strategy in our data pipeline. Initially, the V2X-Sim dataset will be
used in a simulation environment to extract multi-agent sensor recordings from multiple
vehicles. Simultaneously, we will leverage the CIC-IDS-2017 dataset for evaluating network-
level intrusion detection.

Intrusion Detection Systems (IDSs) and Intrusion Prevention Systems (IPSs) remain critical
defense mechanisms against the growing sophistication of network attacks. However, anomaly-
based IDSs often face challenges due to the limited availability of reliable, diverse datasets,
which affects consistent and accurate performance evaluation.

To overcome this limitation, our primary testbed will be the open-source V2X-Sim platform,
enabling low-latency simulation and real-time data extraction. This setup allows our system to
remain flexible and adaptive in various scenarios. Additionally, to further enrich the training data,
we plan to use Generative Adversarial Networks (GANs) to create synthetic attack scenarios,
enhancing our model’s robustness against unseen threats.

This research aims to tackle these security challenges by developing an intrusion detection
system (IDS) powered by deep learning, specifically LSTM (Long Short-Term Memory) and
Autoencoders. The goal is to detect anomalous or malicious V2V messages in real-time,
ensuring that only authentic and trustworthy information is shared among vehicles, thereby
enhancing the security and reliability of autonomous and connected transportation systems.

2�2 Benefits of use case

This research strengthens the foundation of intelligent transport systems by focusing on
cybersecurity in vehicle-to-vehicle (V2V) communication, an essential component of modern
connected infrastructure. Intelligent transport networks rely heavily on accurate and timely
data exchange between vehicles, but they face serious threats from cyberattacks such as false
data injection, GPS spoofing, and denial-of-service (DoS) attacks. These can lead to major
disruptions, including traffic accidents and systemic failures. By addressing these vulnerabilities,
the research helps safeguard critical infrastructure and ensure the safe operation of smart
transportation systems.
A secure V2V communication network is also vital for sustainable urban mobility. Manipulated
or hacked data can trigger false alerts, misdirect traffic, or cause unnecessary congestion—all
of which compromise public safety and efficiency. This work ensures that connected vehicles
share only trustworthy, verified information, thus improving traffic flow, reducing the likelihood
of accidents, and enabling safer autonomous driving environments.
Moreover, the research supports broader efforts to strengthen digital governance in
transportation. Cyber threats in this space not only affect physical safety but also challenge public
trust and institutional control. By developing real-time cyberattack detection and mitigation
methods, this work contributes to the protection of digital infrastructure and aligns with global

790