Name Entity Recognition (NER) Based Drug Related Page Classification on Dark Web

While researching the dark web marketplaces, it was observed that the drugs’ names varied on different marketplaces. Therefore, the same drug might refer to different names on different dark web marketplaces. For example, some marketplaces use the chemical name or medical name as their product name to ensure the exact product; meanwhile, some marketplaces use the street name as the product name to attract users and get more orders. The present work discussed a NER based method to find if a website on the dark web has mentioned drugs. First, the dark web crawler crawled data from the dark web. Then, the authors introduced the Named Entity Recognition (NER) drug dataset with two categories of drug-named entities: Street name and Chemical name. Further, to identify drug-related web pages comprising street and chemical names of drugs with the NER model employed on scraped data. The proposed NER model was tested with the Drug-NER dataset. The DrugcrossNER project contains a predefined Drug-NER dataset with over 3500 listings from the dark web markets. The proposed work also generates a DRUG entity for the NER model in spaCy, an open-source NLP library in python, as it does not have the in-built ability to classify objects into custom categories.