Published Paper Details:

The continuous provision of services in organizations through the IoT (Internet of Things) has made it a new gateway for cyber-attacks. The risks of software piracy and malware attacks are high, which can lead to economic and reputational damages due to the theft of important information. Cybercrime is a growing concern in the cyber age, and classification methods like support vector machine (SVM) and K-nearest neighbour (KNN) are used to classify cybercrime data. Unsupervised classification methods include K-means clustering, Gaussian mixture model, and cluster quasi-random via fuzzy C-means and fuzzy clustering. Neural networks are used to determine synthetic identity theft. Cybercrime detection uses datasets from CBS open data StatLine, with personal characteristics of crime victims. Different training and testing data are analysed for performance. The best technique is used to identify criminals, and the Gaussian mixture model in the unsupervised method shows enhanced performance. The accuracy of the detection method is 76.56%, while the SVM classifier achieves 89% accuracy. Performance metrics include true positive, false positive, false negative, false alarm rate, detection rate, accuracy, recall, precision, specificity, sensitivity, and Fowlkes-Mallows scores. The expectation-maximization (EM) algorithm is used to assess the performance of the Gaussian mixture model. In this paper, we present a distributed framework based on deep learning that can detect and classify malicious traffic to ensure the security of IoT systems. We evaluate two different DL models, feed forward neural network and long short-term memory, using two different datasets (NSL-KDD and BoT-IoT) in terms of performance and identification of different kinds of attacks.