Effective DDoS attack detection in software-defined vehicular networks using statistical flow analysis and machine...

Area of Science:

• Cybersecurity
• Network Engineering
• Artificial Intelligence

Background:

• Vehicular Networks (VN) are crucial for traffic optimization and safety.
• Software-Defined Networking (SDN) enhances wireless network capabilities.
• VN are increasingly vulnerable to Distributed Denial of Service (DDoS) attacks.

Purpose of the Study:

• To propose novel methodologies for detecting DDoS attacks in Software-Defined Vehicular Networks (SDVN).
• To implement Machine Learning (ML) algorithms within SDN Intrusion Detection Systems (IDS) for vehicular environments.
• To address challenges of imbalanced datasets and distinguish between different attack types.

Main Methods:

• Statistical flow analysis and entropy computation.
• Implementation of ML algorithms (K-nearest Neighbor, Random Forest, Logistic Regression) on the BoT-IoT dataset.
• Feature subset selection to optimize model accuracy and evaluate dataset attribute impact.

Main Results:

• The Random Forest classifier achieved high performance metrics: 92% Precision, 92% F1-score, 91% Accuracy, and 90% Recall over five iterations.
• The study identified optimal sample sizes and evaluated dataset attribute impacts on performance.
• The proposed methodology effectively distinguishes between reconnaissance, DoS, and DDoS traffic.

Conclusions:

• Machine Learning, particularly Random Forest, is effective for DDoS detection in SDVN.
• Efficient data handling and potential edge computing are crucial for real-time performance.
• The developed approach offers a scalable solution for enhancing vehicular network security.