Deep Learning Based Artificial Neural Network: An Approach For For Classification Of Attacks In Iot Based Ststems

Abstract

The Internet of Things (IoT) refers to the network of smart heterogeneous objects or devices. The objects in IoT are heterogeneous because they are capable of communicating with other objects or devices. Now-a-days, the adoption of IoT is growing exponentially because of their ability to yield better services. The devices in IoT are resources constrained and are prone to various types of attacks. The attacks in IoT are possible by launching the vulnerabilities in devices, associated software and the network infrastructure. The attacks can be launched locally or globally in the network. There are various methods to classify the attacks. Deep Learning is a subset of machine learning, that is capable of learning unsupervised data .The data in the unsupervised learning are unstructured or unlabelled. They are also referred to as Deep Neural Learning or Deep Neural Network. The proposed Deep Learning based Artificial Neural Network is one among the advanced methods to classify the attacks. It has been analysed using a KDD CUP 1999 dataset containing information about various types of attacks. The attacks are launched by the hackers on the supported features. The info set are obtained from UCI machine learning repository. The activation functions used in this paper are rectifier and softmax activation function. The experimental results of intrusion classification using Artificial Neural Network and Gated Recurrent Unit are compared to find the accuracy in detection rate.