Search Results

The Increasing Prevalence Of Portable Document Format (PDF) Files In Everyday Digital Communication Has Also Made Them A Prime Target For Cyberattacks, As Adversaries Exploit Embedded Scripts, Obfuscated Content, And Structural Vulnerabilities To Deliver Malware. Traditional Signature-based Detection Methods Often Fail Against Novel Or Obfuscated Malicious PDFs, Necessitating More Intelligent And Adaptive Approaches. Machine Learning (ML) Offers A Promising Solution By Learning Discriminative Patterns From Benign And Malicious Files, Enabling Detection Beyond Known Signatures. This Work Proposes A Malicious PDF Detection Framework Based On Machine Learning With An Enhanced Feature Set That Integrates Structural Properties, Content-based Attributes, Metadata Analysis, And Behavioral Indicators. By Combining Static And Dynamic Feature Categories, The Framework Improves Robustness Against Evasion Techniques While Maintaining Scalability. Extensive Evaluation On Benchmark Datasets Demonstrates That The Enhanced Feature Representation Significantly Boosts Classification Accuracy, Reduces False Positives, And Provides Better Generalization Compared To Conventional Feature-limited Models. The Proposed Approach Highlights The Effectiveness Of Enriched Feature Engineering In Strengthening Machine-learning–based Defenses Against Evolving Malicious PDF Threats.