A novel adaptive network intrusion detection system for internet of things

Aravamudhan, Parthiban; Kanimozhi, T.

doi:10.1371/journal.pone.0283725

Cited by 14 publications

(1 citation statement)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, we understand the importance of developing IoT-specific datasets. Many intrusion/malware detection algorithms exist in the literature as described by Thakkar and Lohiya (2021); Asharf et al (2020); Banaamah and Ahmad (2022); Alzubi et al (2022;2023). Table 2 exhibits a summary of a few different studies conducted specifically on IDS for IoT so far.…”

Section: Problem Statementmentioning

confidence: 99%

Generating IoT Specific Anomaly Datasets Using Cooja Simulator (Contiki-OS) and Performance Evaluation of Deep Learning Model Coupled with Aquila Optimizer

Choudhary,

Tanwar,

Choudhury

2024

Journal of Computer Science

View full text Add to dashboard Cite

In recent times, the massive expansion of the Internet of Things (IoT) has transformed various facets of everyday life and industries. The compelling cause behind the widespread adoption of IoT is the increasing availability of affordable, compact, and energyefficient computing devices. While these devices offer significant benefits, they also raise substantial security and privacy challenges. Consequently, safeguarding IoT networks and devices is imperative. To raise a robust security system for IoT networks, it is crucial to have an efficient anomaly-based intrusion detection system. In this study, we introduce a meticulous methodology to create IoT-specific datasets. Utilizing the Contiki-OS Cooja simulator, we generate datasets representative of real-world IoT security threats, including sinkholes, version numbers, and flooding attacks. We then evaluate the performance of a Convolutional Neural Network paired with an Aquila Optimizer (CNN-AO) using these self-generated datasets, by employing metrics such as accuracy, precision, recall, F1-score, sensitivity, specificity, and false alarm rate. Additionally, we compare the effectiveness of CNN and LSTM models in distinguishing between benign and malicious traffic. Our findings demonstrate that the CNN-AO model surpasses other models in accurately classifying normal and malicious traffic with an accuracy of 99.22, 99.77, and 99.55% for our self-generated malicious datasets based on sinkhole attack, version number attack, and flooding attack respectively. This novel approach not only establishes a solid foundation for future investigations in this domain but also provides valuable insights into enhancing IoT system security. In this study, we contribute to the field by introducing a robust methodology for IoTspecific dataset generation and evaluating a cutting-edge CNN-AO model for intrusion detection. Furthermore, it is crucial to note that this research was conducted with utmost ethical consideration. Ethical guidelines and data privacy concerns were meticulously addressed during the generation of IoT datasets and the simulation of real-world attack scenarios, ensuring the responsible conduct of our study.

show abstract