Intrusion Detection in IoT Networks Using Hyperdimensional Computing: A Case Study on the NSL-KDD Dataset

TL;DR

Problem: Intrusion detection in IoT networks with high-dimensional, heterogeneous data. Approach: an HDC-based classifier that encodes each network sample into a $D$-dimensional binary hypervector using base and value hypervectors, then classifies via cosine similarity to learned class centroids. Contributions: a detailed HD encoding scheme with $D=10{,}000$-dimensional hypervectors, binarization, and iterative centroid refinement enabling discrimination among DoS, Probe, R2L, U2R, and normal traffic. Findings: evaluated on NSL-KDD, the method achieves $99.54\%$ accuracy and outperforms several baselines while maintaining computational efficiency. Impact: suitable for resource-constrained IoT deployments and adaptable for online/extended intrusion detection scenarios.

Abstract

The rapid expansion of Internet of Things (IoT) networks has introduced new security challenges, necessitating efficient and reliable methods for intrusion detection. In this study, a detection framework based on hyperdimensional computing (HDC) is proposed to identify and classify network intrusions using the NSL-KDD dataset, a standard benchmark for intrusion detection systems. By leveraging the capabilities of HDC, including high-dimensional representation and efficient computation, the proposed approach effectively distinguishes various attack categories such as DoS, probe, R2L, and U2R, while accurately identifying normal traffic patterns. Comprehensive evaluations demonstrate that the proposed method achieves an accuracy of 99.54%, significantly outperforming conventional intrusion detection techniques, making it a promising solution for IoT network security. This work emphasizes the critical role of robust and precise intrusion detection in safeguarding IoT systems against evolving cyber threats.

Figures (1)

• Figure 1: Overview of the proposed hyperdimensional computing (HDC)-based intrusion detection framework. The process begins with feature encoding, where each network sample’s attributes are transformed into high-dimensional binary hypervectors. Value hypervectors are generated by flipping bits systematically to encode feature variations. These base and value hypervectors are combined using element-wise operations to create a unique representation for each sample. The binarization step optimizes computational efficiency by converting continuous values into binary format. During training, class-specific hypervectors are iteratively refined using the mean of encoded samples per class, enhancing classification accuracy. Finally, new test samples are classified based on their similarity to learned class representations, distinguishing between normal and attack categories (DoS, probe, R2L, U2R). This framework effectively balances accuracy and computational efficiency, making it suitable for intrusion detection in IoT networks.