Deep Learning Solution for Monitoring Social Media Drug Sales

The illegal sale of drugs using social media is a matter of increasing concern among governments. This study proposes a deep-learning-based method to monitor such activities and identify drug-related content using an intelligent Telegram-based chatbot system. The proposed framework uses NLP to analyse textual messages and CNNs to detect drug-related images. Data collection is performed using Telegram APIs, followed by preprocessing steps to ensure data quality. The system uses a BERT-based model to classify suspicious messages and generate alerts automatically. Additionally, CNN models analyse multimedia posts containing drug-related images. Experimental results show that combining the detection of text and images significantly enhances overall accuracy and recall. Overall, the system monitors and identifies drug-related activities on Telegram in real time and demonstrates how deep-learning-based chatbot technology can support law enforcement agencies in enhancing online safety.