Efficient IoT Malware Detection Technique Using Recurrent Neural Network

Internet of Things (IoT) devices are becoming more prevalent in a variety of businesses and for a variety of reasons (for example, sensing/collecting environmental data in both civilian and military situations). Because of their growing impact on many different uses and their expanding computational and analytical capacities, they are a potential threat victim for malware intended to hack particular IoT gadgets. For that, in this paper, we have proposed a successful Recurrent Neural Network (RNN) for malware detection. RNN is a sort of artificial neural network in which nodes are linked together to create a directed chart with a time sequence. Multiple trials with varied hyperparameter values in this research were trained. Our thorough trials demonstrated when employing RNN for malware categorization, the embedding size is more important than the input size. RNN performance with two different feature vectors was assessed using hyperparameters to validate RNN as an efficient solution for malware detection. Natural Language Processing (NLP) and feature selection are the two feature vectors. A paired t-test was also employed in this paper to see if the findings were meaningful to one another. In comparison to the chi2 feature selection, RNN with NLP attained the maximum AUC value and a reasonable variance. Based on the actual results. The proposed effective malware detection approach proves 99% detection accuracy with NLP techniques and 89% with feature selection techniques, hitherto results with the RNN classifier.