Antivirus applied to JAR malware detection based on runtime behaviors.

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Java vulnerabilities are exploited in 91% of web exploits.
• Current antivirus solutions often rely on reactive, blacklist-based models.
• There is a need for proactive and intelligent malware detection systems.

Purpose of the Study:

• To develop an AI-powered antivirus for effective Java malware detection.
• To investigate machine learning models for distinguishing benign from malicious JAR files.
• To offer a preventive approach to cybersecurity threats in Java applications.

Main Methods:

• Executing suspected JAR samples in a controlled Windows OS environment.
• Statistically monitoring and analyzing 6824 distinct actions performed by JAR files.
• Investigating various initial conditions, learning functions, and antivirus architectures.

Main Results:

• The developed antivirus achieved an average performance of 91.58% in differentiating malware from benign JAR files.
• The system demonstrated the capability for proactive threat identification.
• Different configurations of the AI antivirus were explored to optimize performance.

Conclusions:

• Intelligent antivirus systems can overcome limitations of commercial solutions.
• The proposed method enables preventive JAR malware detection, moving beyond reactive strategies.
• AI and machine learning offer a powerful approach to enhance cybersecurity against Java threats.