Computer Science · Computer Networks and Communications
IoT based intrusion detection data analysis using deep learning models
Marwa Baich, Nawal Sael, Touria Hamim
See on the mapAbstract
In both the academic and industrial domains, integration of the internet of things (IoT) is now universally accepted as a significant technical achievement. IoT offers a multitude of security issues despite its many advantages, such as protecting networks and devices, handling resourceconstrained network scenarios, and controlling threats to IoT networks. This article gives a state-of-the-art analysis on the application of multiple deep learning (DL) algorithms in IoT intrusion detection systems (IDS), covering the years 2020 to 2024. Moreover, two popular network datasets, NSL-KDD and UNSW-NB15, are used for an experimental evaluation. The study thoroughly examines and assesses the advantages of well-known deep learning algorithms, including DNN, CNN, RNN, LSTM, and FFDNN. The study demonstrates the exceptional performance of the DNN approach on both datasets, with 99.14% accuracy in multiclass classification in NSLKDD and 99.36% accuracy in binary classification. Furthermore, on UNSWNB15, 82.26% of multiclass classifications and 93.96% of binary classifications with a 42-second minimum running time were achieved, along with an excellent performance in reducing false alarms at a rate of 2.19%.