A Survey on Static and Dynamic Analysis ways for Software Security

Software defect prediction is a crucial element of Software Quality Assurance (SQA), aimed at relating defect-prone modules beforehand in the development lifecycle. With the rapid growth in software complexity, traditional defect discovery approaches such as manual testing and rule-based systems have become inadequate. Machine learning (ML) techniques offer a data-driven result by analyzing historical software data and predicting implicit defects with high accuracy. This survey provides a comprehensive review of machine learning techniques for software defect prediction, including traditional models, ensemble learning, deep learning, and transformer-based approaches. It also examines datasets, feature engineering strategies, evaluation criteria, and cross-project prediction styles. Likewise, the paper discusses challenges such as class imbalance, data quality issues, and model interpretability, along with emerging trends like explainable AI and large language models. The study concludes that ML-based defect prediction significantly enhances software quality, although further exploration is needed to improve generalization and transparency.