Edge-Based Detection and Classification of Malicious Contents in Tor Darknet Using Machine Learning

Abstract

With the increase of data in the network, the load of servers and communication links becomes heavier and heavier. Edge computing can alleviate this problem. Due to a sea of malicious contents in Darknet, it is of high research value to combine edge computing with content detection and analysis. Therefore, this paper illustrates an intelligent classification system based on machine learning and Scrapy that can detect and judge fleetly categories of services with malicious contents. Because of the nondisclosure and short survival time of Tor Darknet domain names, obtaining uniform resource locators (URLs) and resources of the network is challenging. In this paper, we focus on a network based on the Onion Router (tor) anonymous communication system. We designed a crawler program to obtain the contents of the Tor network and label them into six classes. We also construct a dataset which contains URLs, categories, and keywords. Edge computing is used to judge the category of websites. The accuracy of the classifier based on a machine learning algorithm is as high as 89%. The classifier will be used in an operational system which can help researchers quickly obtain malicious contents and categorize hidden services.