Cyber-Physical Systems (CPS) are integral to critical infrastructure, necessitating robust and real-time Intrusion Detection Systems (IDS) to safeguard them from emerging threats. However, the resource constraints of CPS devices present significant challenges in deploying traditional deep learning models for intrusion detection. This paper proposes an efficient Convolutional Neural Network (CNN)-based IDS tailored for resource-constrained CPS environments. The model leverages pruning and quantization techniques to reduce its computational footprint while maintaining high detection accuracy. We begin by designing a CNN architecture specifically for the intrusion detection task, trained on a widely-used dataset. To address the limitations imposed by resource constraints, we apply structured pruning to iteratively eliminate redundant parameters, followed by quantization to compress the model further without significant loss of performance. The optimized model is evaluated against standard metrics, demonstrating a significant reduction in model size and inference time with minimal impact on accuracy. Our experimental results indicate that the proposed approach effectively balances the trade-off between model efficiency and detection capability, making it well-suited for real-time deployment in CPS environments. This study contributes to the ongoing efforts to develop lightweight, yet powerful, IDS solutions for the next generation of CPS security.