Software Systems Security: A Systematic Survey of Vulnerabilities, Detection Techniques, and Mitigation Strategies

Software systems remain vulnerable to a wide range of security flaws that arise throughout the development lifecycle. This paper conducts a systematic survey of software systems security focused on three core dimensions: vulnerability classification, detection techniques, and mitigation strategies. Following established systematic review guidelines, relevant studies published in leading journals were identified, screened, and synthesized. Vulnerabilities are organized according to established taxonomies such as the Common Weakness Enumeration, covering memory corruption, injection, authentication failures, and configuration issues. Detection approaches are examined across static analysis, dynamic analysis, hybrid methods, and data-driven techniques including classical machine learning, graph neural networks, and transformer-based models. Mitigation strategies are reviewed from secure coding practices and automated repair to runtime protections and prioritization frameworks. The synthesis shows that while deep learning methods achieve strong results on curated benchmarks, their effectiveness declines under realistic conditions due to data imbalance, limited inter-procedural analysis, and poor generalization. Mitigation techniques often lag behind detection advances, particularly for zero-day threats and large-scale deployment. The survey concludes by highlighting persistent challenges in dataset quality, explainability, and the integration of formal and learning-based methods, offering a structured foundation for future research in software systems security.

کلیدواژه ها:

Software systems security ، vulnerability classification ، vulnerability detection ، mitigation strategies ، systematic survey ، deep learning