The escalating rate of cyberattacks necessitates intelligent defense systems that can adapt their behavior to respond to emerging threats. Traditional signature-based security systems have their limitations in detecting zero-day attacks and exhibit high rates of false positives. In this paper, we propose a novel combined model that leverages CNNs (Convolutional Neural Networks) and RNNs (Recurrent Neural Networks) for detecting cybersecurity attacks. This model is designed with ideas from feature selection and genetic algorithms. Using the NSL-KDD dataset, our proposed ensemble model attains 97.1% accuracy with sub-second inference time and outperforms the previous state of the art by 3.3%. The computational overhead is decreased by 42% using genetic algorithm optimization, and the detection quality is preserved. Qualitative and quantitative testing reveal that the model is capable of real-time deployment in production, achieving false-positive rates of less than 5% and detection times of under 10 s. This research aligns with the advancement of intelligent cybersecurity systems by providing an efficient, resilient, and scalable mechanism for intrusion detection.