Software-Defined Networking (SDN) is a transformative paradigm that enhances network flexibility and scalability but also introduces new security vulnerabilities due to its centralized control plane and dynamic architecture. Intrusion Detection Systems (IDS) play a critical role in defending SDN environments against sophisticated cyber threats. However, traditional IDS models struggle with imbalanced attack datasets, leading to poor detection of minority attack classes while consuming high computational resources. To address these challenges, we propose IDCNN, a novel deep learning-based IDS tailored for SDN security. IDCNN incorporates the Synthetic Minority Over-sampling Technique (SMOTE) to balance attack class distributions, improving detection accuracy for underrepresented threats. Experimental evaluations on the UNSW-NB15 dataset demonstrate that IDCNN achieves a macro-average accuracy of 97% and an overall accuracy of 99.98%, significantly enhancing intrusion detection performance in SDN environments.