Optimized deep stacked autoencoder for ransomware detection using blockchain network

Abstract

The innovation of technologies has become ubiquitous and imperative in day-to-day lives. Malware is the major threat to the network, and Ransomware is a special and harmful type of malware. Ransomware led to huge data losses and induced huge economic costs. Moreover, Ransomware detection is a crucial task to minimize analyst’s workloads. This paper devises a novel deep learning method for detecting Ransomware using the blockchain network. Here, the sequence-based statistical feature extraction is performed, wherein the features are extracted using 2-gram and 3-gram opcodes. Also, the term frequency-inverse document frequency (TF-IDF) is discovered for each feature. Then the Box-Cox transformation is applied to transformation to the data for improved analysis. Also, the feature fusion is progressed using a fractional concept. Finally, the classification of Ransomware is done using Deep stacked Auto-encoder (Deep SAE), wherein the proposed Water wave-based Moth Flame optimization (WMFO) is adapted for generating the optimal weights. The WMFO is designed by integrating Water wave optimization (WWO) and Moth Flame optimization (MFO). The proposed WMFO-Deep SAE outperformed other methods with maximal accuracy of 96.925%, sensitivity of 96.900%, and specificity of 97.920%.