- Home
- Archives
- 2026
- Vol. 5 Issue 3
- Article
3DVI: A Multi-Tier Polymorphic Malware Detection and Application Vulnerability Testing Platform
Authors
1 2 3 4 Student, Department of Computer Science and Engineering, College of Engineering, Adoor, Kerala, India. 5 Assistant Professor, Department of Computer Science and Engineering, College of Engineering, Adoor, Kerala, India.
Article Information
- Published In
- Volume 5, Issue 3 (2026)
- Published Online
- September-December 2026
- Pages
- 918-925
- Article Type
- Review Article
Abstract
View PDFThe rapid growth of software and digital services has increased both the number and complexity of software vulnerabilities and malicious artifacts. Signature-based scanners remain fast and effective for known threats, but their dependence on previously catalogued patterns limits their ability to recognize zero-day, packed, polymorphic, and metamorphic samples. Re-search has therefore explored structural static analysis, entropy measurement, fuzzing, symbolic and concolic execution, graph representations, machine learning (ML), deep learning (DL), cloud reputation, and large language models (LLMs). This survey organizes that literature through a five-dimensional taxonomy covering analysis visibility, detection knowledge, program repre-sentation, execution strategy, and operational response. Repre-sentative systems are compared by input scope, methodology, reported performance, explainability, latency, and limitations. The survey then uses 3-DVI—Dynamic Virus and AI Threat Inspection—as an integrative reference framework that combines ClamAV, Portable Executable (PE) heuristics, Shannon entropy, metamorphic-opcode checks, YARA rules, VirusTotal correlation, generative-AI reporting, and Wazuh SIEM/XDR integration. The framework preserves deterministic detection as the pri-mary evidence source while using generative AI to contextualize findings and recommend analyst-approved mitigation. Twenty-one synthetic unit and integration cases verify the implemented paths, although they do not establish real-world detection ac-curacy. The review identifies persistent challenges in dataset quality, cross-format support, threshold calibration, adversarial robustness, privacy, explainability, and reproducible evaluation, and proposes directions toward lightweight emulation, evidence-grounded LLM analysis, calibrated benchmarking, and scalable deployment.
Related Articles
Share Article
*Instagram doesn't support direct link sharing from web. Copy the link and share it in your Instagram story or post.