SafeDocs: A Machine Learning-Based Framework for Malicious PDF Detection Tailored for SMEs

Cyber threats targeting small and medium enterprises (SMEs) are rising, with weaponized PDF files as prime attack vectors. However, many advanced malicious PDF detection solutions are unfeasible for resource-constrained SMEs. This paper proposes SafeDocs, an automated framework leveraging machine le...

Full description

Saved in:
Bibliographic Details
Published in2023 RIVF International Conference on Computing and Communication Technologies (RIVF) pp. 295 - 300
Main Authors Duc Le, Tran, Le-Dinh, Thang, Uwizeyemungu, Sylvestre, Hung Doan, Xuan, Duy Dinh, Truong
Format Conference Proceeding
LanguageEnglish
Published IEEE 23.12.2023
Subjects
Analytical models
Computational modeling
Computer viruses
machine learning
malicious PDF
malware
Metadata
Portable document format
Prototypes
SME
Weapons
Online AccessGet full text

Cover

More Information
Summary:Cyber threats targeting small and medium enterprises (SMEs) are rising, with weaponized PDF files as prime attack vectors. However, many advanced malicious PDF detection solutions are unfeasible for resource-constrained SMEs. This paper proposes SafeDocs, an automated framework leveraging machine learning to identify malicious PDFs designed specifically for SMEs. SafeDocs analyzes the structural metadata of PDFs using a well-known Random Forest model to classify malicious files. It provides SMEs with an effective and usable solution to counter evolving PDF threats. Experiments demonstrate SafeDocs achieving over 98% accuracy in detecting thousands of malicious PDF samples. A prototype is implemented, and guidelines are provided for seamless integration within typical SME IT infrastructure. Overall, SafeDocs offers SMEs a robust defense against malicious PDFs within their constraints.
ISSN:2473-0130
DOI:10.1109/RIVF60135.2023.10471850