A COMPREHENSIVE SURVEY OF ADAPTIVE RANDOM TESTING AND HYBRID APPROACHES IN CYBERSECURITY SOFTWARE TESTING

Abstract

Cybersecurity software testing has shifted from basic, random approaches to advanced, intelligent frameworks that can identify complex vulnerabilities in modern, distributed systems. This survey offers a thorough review of testing methodologies, focusing on Adaptive Random Testing (ART) and Partition-Based ART (PB-ART) as key techniques that connect traditional and smart testing approaches. It traces the evolution of methods from foundational techniques such as fuzzing, mutation testing, and vulnerability assessment platforms to more sophisticated, hybrid approaches that leverage machine learning for prioritization, clustering, anomaly detection, and optimization. The review also examines specialized frameworks and applications in areas such as IoT, cloud computing, API security, and DevSecOps. Analysing strengths, limitations, and performance metrics highlights research gaps, including input-space explosion, benchmark realism issues, machine learning reliability, and automation challenges. The three-phase framework captures how testing methods have progressed from simple input strategies to highly adaptive systems capable of managing complex cybersecurity environments. Overall, the survey consolidates current knowledge, offers insights across various testing contexts, and suggests future directions for more effective and automated vulnerability detection in critical software systems.