Elevating Smart Industry Security: An Advanced IoT-Integrated Framework for Detecting Suspicious Activities using ELM and LSTM Networks

Abstract

The proliferation of Internet of Things (IoT) devices in smart industrial contexts necessitates robust security measures to thwart potential threats. This study addresses the escalating security challenges arising from the widespread deployment of IoT devices in smart industrial environments. Focusing on the identification and categorization of potentially harmful activities, our research introduces an innovative framework that seamlessly integrates networks of Extreme Learning Machines (ELM) with Long Short-Term Memory (LSTM). The primary goal is to significantly enhance the accuracy and efficiency of real-time detection of suspicious activities. Implemented using Python, the framework exhibits a remarkable 97.5% improvement in recognizing and accurately categorizing suspicious activities compared to traditional methods such as Conv 1D and 3D CNN. Rigorous testing on a substantial real-world dataset simulating smart industry scenarios underlines this substantial improvement over conventional approaches in identifying and precisely classifying questionable activities. The design excels in comprehending complex behavioral trends within the dynamic IoT data environment, leveraging the temporal memory retention capacity of LSTM networks. This research lays the groundwork for fortifying cybersecurity in smart industries against emerging online threats and malicious actions. The proposed framework capitalizes on the synergies between LSTM and ELM networks to achieve heightened accuracy in identifying suspicious activities, providing comprehensive and dynamic insights from real-time IoT data. These insights are crucial for proactive threat detection and prevention in smart industrial settings, contributing to an elevated level of security against evolving threats.