Published Paper Details:

There are many sensors and actuators that make up the IoT, all of which are linked together over wired or wireless networks. Recent years have seen a meteoric rise in the variety of IoT use cases, from smart homes and VANETs to healthcare and smart cities to wearables. According to IHS Markit 1, the number of connected devices is expected to increase from 27 billion in 2017 to 125 billion in 2030, an average yearly increment of 12 percent. Due to the diversity of IoT architecture, devices, and communication channels (most notably wireless) and the sheer volume of data being transferred over these channels, security has become an urgent concern in the IoT space. As the number of IoT-enabled gadgets grows, so does the significance of ensuring their safety. We propose a novel Deep Learning-based Intrusion Detection System (DLIDS) to identify security threats in IoT environments and thereby help address the aforementioned problems with protecting IoT devices. The accuracy of attack detection is negatively impacted by the fact that many IDSs in the literature do not have optimal feature learning and dataset management. Many signature-based and machine learning/deep learning AUTO ENCODER-based CNN algorithms have been proposed to try and detect such attacks, but their success rate has been less than stellar. We introduce Deep Learning IDS based on the Convolution2D Deep Learning algorithm, which employs many layers to filter IOT data and, as a consequence, obtains optimised features that enable the detection of IOT attacks with a prediction accuracy of 99%.