Authors:
Vamshidhar Reddy Vemula

Addresses:
Department of Information Technology, Intalent LLC, Plano, Texas, United States of America.

Abstract:

Geometric expansion of illegal trade on the dark web is a massive threat to international cybersecurity. Passive, non-proactive traditional investigation falls short in facing the dynamic, anonymous nature of dark web markets. The novel VIGILANTE, a fresh paradigm, is proposed here as an AI system for the early detection of such markets before they emerge and the prediction of new cybercrime trends. The study utilizes a pre-filtered dataset, “DarkNet-2M Listings,” which comprises the five most visited (now closed) dark web markets, featuring 2 million web-scraped products gathered over 36 months. Our daily activity is a hybrid AI method. We use Natural Language Processing (NLP) methods, specifically a fine-tuned BERT model, for category tagging and semantic reasoning of illicit products and services. For the forecasting component, a Long Short-Term Memory (LSTM) neural network is subsequently trained on forecasted market movement trends using predicted time-series data from ads. The primary tool used for this work is Python, and the secondary tools utilized with it include the Scrapy library for web scraping, TensorFlow and Keras for model development, and Matplotlib for data visualization. Our result achieves highly precise illegal listing detection and highly precise trend forecasting—observed cybercrime trend matching—thus realizing the potential of AI as an extremely effective tool for security officials and police administrations to transition from a reactive security function to a proactive one.

Keywords: Cybercrime Trends; Long and Short-Term Memory; Natural Language Processing; BERT Model; Tensor Flow and Keras; Artificial Intelligence; Trend Matching; Illegal Trade.

Received on: 12/11/2024, Revised on: 18/01/2025, Accepted on: 27/02/2025, Published on: 05/09/2025

DOI: 10.69888/FTSCL.2025.000428

FMDB Transactions on Sustainable Computer Letters, 2025 Vol. 3 No. 3, Pages: 126-135

• Views : 138
• Downloads : 41

Download PDF