FL-DSFA: Securing RPL-Based IoT Networks against Selective Forwarding Attacks Using Federated Learning

Rabia Khan ¹, Noshina Tariq ¹, Muhammad Ashraf ²

⁰Department of Avionics Engineering, Air University, Islamabad 44000, Pakistan.

Sensors (basel, Switzerland) +

|

September 14, 2024

View abstract on PubMed

Summary

This study introduces FL-DSFA, a federated learning technique to detect selective forwarding attacks in the Internet of Things (IoT). It enhances IoT security by accurately identifying routing attacks with high precision and efficiency.

Area Of Science

Computer Science
Cybersecurity
Machine Learning

Background

The Internet of Things (IoT) enables seamless device integration but faces significant security challenges.
Routing attacks, particularly selective forwarding attacks on the Routing Protocol for Low-Power and Lossy Networks (RPL), pose a severe threat to IoT systems.
These attacks target critical components like control messages and routing topologies, leading to potential data loss and system damage.

Purpose Of The Study

To develop and evaluate a novel federated learning-based detection technique (FL-DSFA) for identifying selective forwarding attacks in IoT networks.
To enhance the security and reliability of IoT routing protocols against sophisticated attacks.
To improve the efficiency and privacy of attack detection mechanisms in resource-constrained IoT environments.

Main Methods

Implementation of a lightweight federated learning model (FL-DSFA) utilizing the IoT Routing Attack Dataset (IRAD).
Inclusion of attack types such as Hello Flood (HF), Decreased Rank (DR), and Version Number (VN) for comprehensive detection.
Assessment of binary classification algorithms including Logistic Regression (LR), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Naive Bayes (NB) for training efficiency.

Main Results

The FL-DSFA technique, particularly with SVM and KNN classifiers, demonstrated high accuracy and efficient runtime performance during training.
The proposed system achieved exceptional performance metrics: 97.50% prediction precision, 95% accuracy, 98.33% recall rate, and 97.01% F1 score.
Comparative analysis confirmed the superiority of the proposed method over existing research in terms of classification accuracy, scalability, and privacy.

Conclusions

Federated learning offers a robust solution for detecting selective forwarding attacks in IoT networks.
The FL-DSFA technique provides a scalable, efficient, and privacy-preserving approach to bolster IoT security.
The study highlights the potential of machine learning in securing critical IoT infrastructure against evolving cyber threats.

Keywords:

Decreased Rank (DR) Hello Food (HF) Internet of Things (IoT) IoT Routing Attack Dataset (IRAD) Linear Discriminant Analysis (LDA) deep learning federated learning