Unveiling malicious PDF behavior: Interpretable classification and profiling malicious PDF using TabNet

doi:10.1016/j.jisa.2026.104487

Unveiling malicious PDF behavior: Interpretable classification and profiling malicious PDF using TabNet

2026-04-27

0

PRE

AI

A

Arousha Haghighian Roudsari *

A

Arash Habibi Lashkari

W

Woong-Kee Loh

DOI：10.1016/j.jisa.2026.104487

Original

Original request for help

Save

Abstract

En 中文

• Malicious PDF detection using TabNet, offering interpretability with high detection performance. • Reduces feature engineering by leveraging TabNet’s dynamic attention mechanism to identify relevant features for classification. • Comprehensive analysis of TabNet’s interpretability provides insights into local and global feature importance. • Achieves state-of-the-art results on the real-world CIC-Evasive-PDFMal2022 dataset.

Keywords:

Malicious PDF detection

TabNet

Interpretability

Feature importance

Machine learning

Unveiling malicious PDF behavior: Interpretable classification and profiling malicious PDF using TabNet

Abstract

Journal

Organization