Reproduce “CTI Dataset Construction from Telegram” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Dincy R. Arikkat, Sneha B. T., Serena Nicolazzo, Antonino Nocera, Vinod P., Rafidha Rehiman K. A., Karthika R

Researchers from Cochin University of Science and Technology and collaborating Italian universities developed an automated pipeline to construct a high-fidelity Cyber Threat Intelligence (CTI) dataset from Telegram. The system collected 145,349 messages, used a BERT-based model achieving 96.64% accuracy to filter relevant content, and curated 86,509 verified malicious Indicators of Compromise (IoCs).

Key insights

  • Telegram represents a significant and underutilized source of real-time Cyber Threat Intelligence, offering early signals of vulnerabilities and threats.
  • Fine-tuned BERT models demonstrate strong performance (96.64% accuracy) in filtering relevant CTI from the high volume of unstructured data found in messaging platforms.
  • A multi-step verification process, involving cross-referencing with public databases, is essential to confirm the maliciousness of extracted IoCs, particularly for non-CVE types.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.