The rapid expansion of cloud computing, Internet of Things (IoT), and large-scale distributed systems has significantly increased network complexity and exposure to cyber threats. Traditional rule-based intrusion detection systems are limited in identifying modern and evolving attacks, while machine learning-based approaches, although highly accurate, often suffer from poor interpretability and reduced long-term reliability due to changing network behavior. This review paper presents a comprehensive analysis of hybrid intrusion detection systems that integrate rule-based intelligence with machine learning models enhanced by explainable artificial intelligence and adaptive learning mechanisms. Key techniques such as gradient boosting classifiers, SHapley Additive exPlanations (SHAP), and incremental learning strategies are examined in the context of improving detection accuracy, transparency, and resilience against concept drift. Widely used benchmark datasets including CICIDS2017, UNSW-NB15, and NSL-KDD are reviewed along with standard performance evaluation metrics. The paper highlights current research trends, practical challenges, and future directions for building trustworthy and scalable cyber defense frameworks suitable for dynamic distributed environments.