With the rapid expansion of network-based applications, intrusion detection systems (IDS) have become essential for safeguarding sensitive data and ensuring system reliability. However, traditional IDS models often face challenges related to high computational cost and energy consumption, especially in resource-constrained environments. This paper presents a comprehensive review of energy-efficient IDS approaches based on machine learning and deep learning techniques. It covers various methods, including CNN, LSTM, auto encoders, optimization algorithms, and hybrid models, highlighting how they improve detection accuracy while reducing computational overhead. The review also analyzes datasets, performance metrics, and energy optimization strategies such as pruning, quantization, and knowledge distillation. By categorizing existing research and identifying gaps, this work aims to provide insights into designing scalable, low-latency, and energy-aware IDS solutions for modern networks. The findings serve as a foundation for future advancements in sustainable and intelligent cyber security systems.