Page 825 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

(continued)

Domain Intelligent Transport
Testbeds or Pilot Pilot deployment on OBUs in real-world vehicles or simulation envi-
Deployments ronments to test the IDS’s real-time performance and accuracy in Transport 4.10: Intelligent
detecting intrusions. Can use vehicle communication simulators (V2X-Sim)
or controlled traffic scenarios.

2 Use Case Description

2�1 Description

The Internet of Things (IoT) has transformed various industries over the years, and now a
new emerging field in transportation safety, known as the Internet of Vehicles (IoV), is
gaining prominence. While the core technology behind IoV is Vehicle-to-Everything (V2X)
communication, where vehicles interact with their surroundings, the necessity for this technology
arises from a pressing global concern—road safety.

Every year, a significant number of fatal road accidents occur, making transportation safety a
top priority. Vehicle-to-Vehicle (V2V) communication is a crucial subset of V2X that enables
vehicles to exchange real-time information to prevent collisions, coordinate movement, and
enhance road safety. However, as with any technological advancement, new vulnerabilities
and security threats emerge. The dataset employed in this research is the V2X-Sim 2.0 dataset,
a multi-sensor, annotated dataset specifically developed for research in autonomous driving
and vehicle-to-everything (V2X) communication. It follows the same schema as the nuScenes
dataset, ensuring a well-structured and comprehensive format for training and evaluating
perception models in V2V environments.
One of the most significant challenges in V2V communication is ensuring the security of
transmitted data against cyber threats. Malicious activities such as false data injection, spoofing,
Sybil attacks, and denial-of-service (DoS) attacks can compromise network integrity. If false
information is introduced into the system, it can mislead vehicles and lead to hazardous
situations, potentially endangering lives. This highlights the importance of robust Intrusion
Detection Systems (IDS) tailored to the unique demands of V2V networks.

For evaluating our IDS, we will use standard classification metrics, including Accuracy, Precision,
Recall, and F1-Score.

In addition to V2X-Sim, we will utilize the CIC-IDS-2017 dataset, a widely used benchmark for
network intrusion detection. This dual-dataset approach supports both perception-level data
from vehicle sensors and network-level traffic data for intrusion analysis.

We employ Long Short-Term Memory (LSTM) networks and Autoencoders as our primary
deep learning models:

1. LSTM is well-suited for modeling time-series data, capturing sequential patterns in
vehicular communication.
2. Autoencoders, as an unsupervised learning method, are effective for anomaly detection,
learning the normal behavior of network transmissions, and identifying outliers or
deviations.

789