Unveiling the dark: Analyzing and Categorizing dark web activities using Bi-Directional LSTMs

The term "dark web" refers to a broad range of illegal actions which are performed by unidentified individuals or groups, making it hard to identify their origins. The illicit material discovered on the dark web undergoes is frequently updated and altered, posing a significant challenge in detecting and categorizing these unlawful activities. Addressing this issue has recently become a pressing matter that demands prompt attention from both professionals in the industry and experts in academia. To tackle this problem, a document is presented that introduces a specialized web crawler capable of gathering, cleansing, and storing dark web pages in document databases. The web crawler automatically categorizes the collected web pages into five distinct groups. To classify these pages, classifiers such as Support Vector Machines (SVM), long-term memory (LSTM), and bidirectional LSTM are utilized. Experimental results indicate that the bidirectional LSTM and SVM classifiers achieve an accuracy of 92% and 81% respectively.