Trust-Aware Five-Class Multinomial Logistic Regression for Packet-Dropping Mitigation in Mobile Ad Hoc Networks
Year: 2026
Type: article
Abstract: Mobile Ad Hoc Networks (MANETs) are highly vulnerable to packet-dropping attacks because of their dynamic, infrastructure-less topology.Existing binary classification and trust-based approaches often fail to distinguish malicious behaviours, such as black hole and gray hole attacks, from non-malicious packet losses caused by buffer overflow and energy depletion, thereby leading to false positives and inefficient routing decisions.This study proposes a trust-aware framework that combines multinomial logistic regression (MLR) for five-class node behaviour classification with a dynamic reputation-based trust model.Features extracted from NS-2 simulations are used to classify node behaviours, and the classification results are incorporated into real-time trust updates for secure Ad hoc On-Demand Distance Vector (AODV) routing.Experimental results show that the proposed method achieves a classification accuracy of 95.7%, with precision and recall ranging from 0.93 to 0.98 across all classes and an area under the curve (AUC) greater than 0.90.In routing evaluation, the framework attains a 95% packet delivery fraction (PDF), representing an 80% improvement over the baseline reactive protocol under misbehaviour and a 31% gain over multifactor mitigation, while maintaining low delay (0.20 s) and low energy consumption (93.27J).These findings indicate that the proposed approach can reduce false positives, improve packet delivery, and enhance routing reliability in resource-constrained MANET environments.
Institution: GITAM University
Language: English
Topic: Mobile Ad Hoc Networks
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7147205237