Author ORCID Identifier
0000-0001-9765-0656
Abstract
Mobile devices and wireless sensor networks (WSNs) are increasingly vulnerable to security threats such as unauthorized access and replica node attacks. Mobile devices face risks from replication and anomalous behavior, while attackers compromise WSNs by cloning legitimate nodes, thus threatening network integrity. Traditional security mechanisms often fall short in detecting such sophisticated threats, especially in resource-constrained environments. This research proposes a dual-component security system. A Machine Learning-Based Intrusion Detection System (IDS) for WSNs leverages Graph Neural Networks (GNNs) to detect replica nodes through structural network analysis and applies Federated Learning to preserve data privacy. The Sequential Probability Ratio Test (SPRT) enables swift anomaly detection, while lightweight algorithms improve energy efficiency.Additionally, the proposed History of Neighbor Node (HNN) technique enhances detection accuracy by recognizing replica nodes at both local and global levels, even under varying mobility constraints. On mobile devices, anomaly detection models such as Isolation Forest (99% accuracy), Support Vector Machine (90%), and Random Forest (86%) monitor user behavior to identify unauthorized access. A real-time alert mechanism notifies authorized personnel of potential threats. Overall, the system presents a scalable, energy-efficient, and high-accuracy approach to safeguard mobile devices and WSNs against dynamic replication-based threats.
Recommended Citation
Pavani, Maram; Sharani, Tanguturi; and S A, Amutha Jeevakumari
(2025)
"A Machine Learning-Driven Framework for Real Time Detection and Prevention of Replica Node Attacks in Wireless Sensor Networks,"
Northeast Journal of Complex Systems (NEJCS): Vol. 7
:
No.
1
, Article 16.
DOI: https://doi.org/10.63562/2577-8439.1102
Available at:
https://orb.binghamton.edu/nejcs/vol7/iss1/16