Page 824 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 8: AI for Intrusion in Vehicle-to-Vehicle (V2V)

Communication

Country: Nigeria

Organization: AI4Africa Research Group

Contact Person(s):

Chidi Ebube (chidizack24@ gmail .com), Dr Houda Chihi (houda.chihi@ supcom .tn), Blessed
Guda (gudablessed@ gmail .com), Emmanual Aaron (aaronemmanuel054@ gmail .com),
Emmanuella Sule (suleemmanuella@ yahoo .com), Emmanuel Ani (ani.mlengineer@ outlook
.com), Okafor Miracle Uche (okaformiracle212@ gmail .com)

1 Use Case Summary Table

Domain Intelligent Transport
Problem to be Detect and prevent intrusions or attacks (e.g., spoofing, jamming, unau-
Addressed thorized data injection) within V2V communication systems to ensure
vehicle safety and coordination.

Key Aspect of the Deep learning-based intrusion detection system (IDS) using LSTM for
Solution sequential anomaly detection and Autoencoders for outlier detection in
V2V communication.

Technology Deep Learning, LSTM, Autoencoders, Intrusion Detection System (IDS),
Keywords Vehicle-to-Vehicle (V2V), OBUs (Onboard Units), Anomaly Detection,
Edge Computing.

Data Availability V2X-Sim Dataset [5]
Network Intrusion Dataset [6]

Metadata (Type of Vehicle communication logs, sensor data, traffic information (speed, loca-
Data) tion, braking, warnings), labeled data (normal/abnormal communication
patterns), timestamps, event types (e.g., collision warnings).

Model Training and - LSTM: Detects anomalies in sequential V2V communication data.
Fine-Tuning - Autoencoders: Learns normal traffic patterns and flag outliers as poten-
tial intrusions.
- Supervised & Unsupervised Learning: Uses labeled attack data where
available and unsupervised methods where labels are missing.

788