The evolution of the Internet of Things (IoT) has various security aspects, such as smart devices being heavily used in resource-poor smart environments, which could be attacked by cyber threats. Intrusion Detection Systems (IDS’s) are vital for detecting threats and cyber-attacks, but traditional security approaches are often too restrictive and inappropriate for IoT devices. This work aimed to identify and apply the most important network traffic attributes to reduce computational complexity and improve detection efficiency. Multiple Machine Learning(ML) and Deep Learning(DL) techniques, including Convolutional Neural Network (CNN), Artificial Neural Network (ANN), Support Vector Machine (SVM), Long Short Term Memory (LSTM), Random Forest (RF) and Decision Tree(DT) are assessed to attack classification. Experimental findings indicate that the RF model attained better performance for the major attack types (DOS_SYN_Hping and MQTT_Publish), moderate performance for DDOS_Slowloris and poor performance for some. Some rare attack classes, such as NMAP_FIN_SCAN and Wipro_bulb.In comparison, Random Forest (RF) and Decision Tree (DT) performed better in terms of accuracy and predicted low latency. Thereby to provide the proposed system, Performance-driven Multi-Model Evaluation Framework (MMEF) for IoT Intrusion Detection using Efficient ML and DL strategies aims to improve cyber-threat detection accuracy and reduce computational overhead to provide an effective and scalable solution for securing IoT network threats against multi-model cyber-attacks.
 

Machine Learning, Deep Learning, Future Selection, Intrusion Detection System, IoT Security and Multi-model Attack Detection