Classification of Dark Web Using Text Based CNN and Topic Weight Model

It is difficult to monitor its users, the Dark Web, an online domain that guarantees user anonymity, has grown to be a hub for illicit activity and a source of information about cyberattacks. This study looked at how the Dark Web is categorised in connection with various online dangers. To identify vector types appropriate for machine learning categorisation, we analysed words from the Dark Web. Conventional techniques that build features by using all Dark Web texts produce vectors that contain every word on the Dark Web. Nevertheless, this method adds unnecessary information to the vectors, which reduces learning efficiency and lengthens processing time. By concentrating on certain keywords within each class, the study sought to reduce the size of the word vectors and improve the categorisation process. Utilising the Dark Web's anonymity feature and topic-modeling-based weight creation made this optimisation possible. These techniques improved the differentiation of Dark Web classes by enabling the construction of word vectors with a limited feature set. We combined TextCNN with topic modelling weights in order to enhance classification performance even further. We used two datasets for validation and evaluated the model's performance against alternative text classification methods; the suggested model outperformed the others in Dark Web categorisation.