Collaborative Learning Based Sybil Attack Detection in Vehicular AD-HOC Networks (VANETS).

Area of Science:

• Computer Science
• Network Security
• Artificial Intelligence

Background:

• Vehicular Ad-hoc Networks (VANETs) face significant security challenges due to their dynamic topology.
• Sybil attacks, which involve forging false identities, pose a critical threat to VANETs, impacting safety and traffic management.
• Existing security strategies are insufficient to fully address the pervasive Sybil attack in VANETs.

Purpose of the Study:

• To propose a novel collaborative framework for detecting Sybil attacks in VANETs.
• To enhance the security and reliability of VANET communications, particularly for safety-critical applications.
• To evaluate the effectiveness of a majority voting mechanism in identifying Sybil attack instances.

Main Methods:

• A collaborative framework was developed, ensembling multiple classifiers including K-Nearest Neighbor, Naïve Bayes, Decision Tree, SVM, and Logistic Regression.
• A majority voting mechanism (both Hard and Soft voting) was employed for final Sybil attack prediction.
• The framework's performance was rigorously evaluated using accuracy metrics and Receiver Operating Characteristic (ROC) curves.

Main Results:

• The proposed collaborative framework achieved a high detection accuracy of 95% for Sybil attacks.
• Both Hard and Soft majority voting mechanisms were compared, providing insights into optimal prediction strategies.
• The Receiver Operating Characteristic (ROC) curve analysis confirmed the framework's effectiveness in distinguishing between normal and malicious nodes.

Conclusions:

• The novel collaborative framework effectively detects Sybil attacks in VANETs with high accuracy.
• Majority voting serves as a robust mechanism for consolidating predictions from diverse classifiers in VANET security.
• The proposed approach significantly contributes to improving the security and trustworthiness of VANETs, crucial for intelligent transportation systems.