Reproduce “DNS Tunneling: A Deep Learning based Lexicographical Detection Approach” 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

Franco Palau, Carlos Catania, Jorge Guerra, Sebastian Garcia, Maria Rigaki

Researchers at Universidad Nacional de Cuyo and Czech Technical University developed a 1D Convolutional Neural Network for lexicographical detection of DNS tunneling, achieving a 96% F1-score and a low 0.52% false positive rate. This work also introduces and publicly releases a new dataset, meticulously generated from multiple DNS tunneling tools, to address the scarcity of evaluation resources.

Key insights

  • A relatively simple 1D-CNN architecture can effectively learn complex lexicographical patterns to distinguish malicious DNS tunneling domains from legitimate ones, enabling efficient detection.
  • The scarcity of high-quality, labeled datasets for DNS tunneling can be successfully addressed through systematic generation in controlled environments, providing a crucial resource for the research community.
  • Achieving a remarkably low False Positive Rate (FPR) is critical for the practical deployment of DNS tunneling detection systems, minimizing alert fatigue in operational security environments.

Summary by alphaXiv.

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