With the rise of deceptive job offers, financial scams, and grooming on platforms like Reddit, safeguarding users against social engineering threats has become increasingly critical. Traditional moderation methods often fall short due to the scale and evolving nature of such attacks. To address this, we present a hybrid Open-Source Intelligence - Natural Language Processing (OSINT-NLP) pipeline that automates the detection of suspicious Reddit posts. High-risk subreddits in this study were defined as communities that have been reported or observed to facilitate activities such as online grooming, human trafficking, exploitation, fraudulent recruitment, and other forms of malicious social engineering. The system collects data via the Python Reddit API Wrapper (PRAW API), targeting high-risk subreddits and keywords. Posts are preprocessed using spaCy and regular expressions to extract entities like emails, phone numbers, and URLs. A rule-based engine provides initial risk scores (Low, Medium, High), which are refined by a fine-tuned DistilBERT classifier. For deeper threat attribution, the pipeline integrates Maigret, enabling the profiling of usernames across over a hundred online platforms. The system is deployed via a Streamlit dashboard for real-time monitoring and a Flask app for structured analysis. Tested on about 5000 posts, it achieved 93.7% accuracy, with significant improvement in High-risk F1 scores. Our results show that blending heuristics, deep learning, and OSINT offers a scalable, explainable solution for automated threat detection and digital safety enforcement.
Paper
Full text
Social Engineering Threat Detection on Reddit Using OSINT and Transformer-Based NLP Model
Semantic Scholar · 2026
Abstract
With the rise of deceptive job offers, financial scams, and grooming on platforms like Reddit, safeguarding users against social engineering threats has become increasingly critical. Traditional moderation methods often fall short due to the scale and evolving nature of such attacks. To address this, we present a hybrid Open-Source Intelligence - Natural Language Processing (OSINT-NLP) pipeline that automates the detection of suspicious Reddit posts. High-risk subreddits in this study were defined as communities that have been reported or observed to facilitate activities such as online grooming, human trafficking, exploitation, fraudulent recruitment, and other forms of malicious social engineering. The system collects data via the Python Reddit API Wrapper (PRAW API), targeting high-risk subreddits and keywords. Posts are preprocessed using spaCy and regular expressions to extract entities like emails, phone numbers, and URLs. A rule-based engine provides initial risk scores (Low, Medium, High), which are refined by a fine-tuned DistilBERT classifier. For deeper threat attribution, the pipeline integrates Maigret, enabling the profiling of usernames across over a hundred online platforms. The system is deployed via a Streamlit dashboard for real-time monitoring and a Flask app for structured analysis. Tested on about 5000 posts, it achieved 93.7% accuracy, with significant improvement in High-risk F1 scores. Our results show that blending heuristics, deep learning, and OSINT offers a scalable, explainable solution for automated threat detection and digital safety enforcement.