<p>The devices connected in Internet-of-Things (IoT) networks are battery-powered, making energy efficiency an important consideration, while secure data transmission is key to protecting sensitive information. Existing solutions cannot adapt to changing network conditions while resulting in wasted energy, reduced throughput, and increased latency. This research presents an innovative framework for optimizing secure data transmission in IoT networks using energy-efficient machine-learning approaches. Different from existing models that consider security and energy efficiency separately, the proposed model presents an integrated framework that co-optimises both in real-time. The model introduces innovativeness by using machine learning for dynamically fine-tuning transmission power and transmission frequency. Finally, the model presents a real-time adaptive security technique that opts for security protocols based on network load, level of device battery, and data sensitivity. The accomplished results from rigorous testing show a 30% reduction in energy usage, 25% increase in throughput, 15% reduction in latency, and 20% improvement in response time when compared to existing traditional approaches. In addition, the machine learning model predicted transmission parameters with 90% accuracy, beating static, heuristic-based techniques.</p>