Multidimensional Forensic Investigation of Onion Sites Based on Fuzzy Encoded LSTM

Abstract

The only way to access onion services is via the TOR browser providing anonymity and privacy to the client as well as the server. Information about these hidden services and the contents available on them cannot be gathered like websites on the surface web. So, they become a fertile ground for illegal content dissemination and hosting for cybercriminals. There is a persistent need to classify and block such content from onion sites. In this paper, we investigate data requested from onion services to help law enforcement agencies collect traces of cybercrime on these hidden services. We propose a system using fuzzy encoded LSTM to analyze contents retrieved from these sites and raise alerts if found illegal. The accuracy of fuzzy-encoded LSTM is found to be 81.04 % and it outperforms other classifiers.