Dark web crawling for automatic dark web classification using Bayesian Hierarchical Neural Attention Harmonic Network

Dark web is the canopy idea that specifies any sort of illegal actions conducted by unknown organizations or persons, thus making it complicated to track. The illegal contents on dark web are changed and updated constantly. The classification and collection of such activities are highly challengeable tasks, since they are time-consuming and difficult. In present times, it is emerged as a problem that needs rapid attention from academia and industry. In this research, Bayesian Hierarchical Neural Attention Harmonic Network (BHNAHN) is presented for dark web classification. Here, dark web crawling and classification of dark web are the two steps conducted. TorBot is employed for dark web crawling based upon keywords like pornography, financial gambling, drugs, hacking, cryptocurrency, arms/weapons, electronics and violence. In dark web classification process, input web data is acquired and then, Bidirectional Encoder Representations from Transformers (BERT) tokenization is carried out. Afterwards, features are extracted from tokenized word. Finally, dark web classification is accomplished employing BHNAHN. However, BHNAHN is modeled by incorporating Bayesian Neural Network (BNN) and Hierarchical Neural Attention classifier with forward harmonic analysis. Additionally, BHNAHN obtained maximal accuracy, True Negative Rate (TNR) and True Positive Rate (TPR) about 91.362%, 92.440% and 90.799