DOI  https://doi.org/10.36719/3104-4735/2/26-29

Huseyn Gasimov

Nakhchivan State University

PhD in Techniques

https://orcid.org/0000-0002-3714-875X

huseynqasimov@ndu.edu.az

Rugayyakhanim Garibzada

Nakhchivan State University

Master’s student

https://orcid.org/0009-0001-4962-918X

ruqeyyexanim202027@icloud

Deep Learning versus Traditional Antivirus Software

Abstract

Cybersecurity has become increasingly complex, with traditional antivirus software struggling to keep up with modern threats such as zero-day exploits and advanced persistent threats (APTs). While traditional methods, including signature-based detection and heuristic analysis, remain effective against known malware, they fall short in detecting new or sophisticated attacks. The rapid evolution of cyberattacks, as well as the emergence of polymorphic and metamorphic malware, severely reduces the effectiveness of traditional defense methods. Because this type of malware changes its code structure with each execution, it easily manages to evade signature-based systems by confusing them. On the other hand, deep learning-based security systems leverage artificial intelligence to analyze patterns and behaviors, providing superior detection of previously unseen threats. These systems analyze large volumes of telemetry data, monitor behavioral changes in real time, and identify subtle patterns that traditional methods cannot detect. It is precisely these capabilities that make them an indispensable component in proactive defense strategies. However, deep learning systems require more computational resources and are vulnerable to adversarial attacks. A hybrid approach that combines traditional antivirus methods with AI-driven solutions offers a promising strategy to enhance cybersecurity defenses, providing comprehensive protection against a wider range of cyber threats.

Keywords: cybersecurity, deep learning, antivirus software, machine learning, threat detection