The advent of smart grids has revolutionized power systems by enabling real-time monitoring, automation, and efficient energy management. However, this interconnected infrastructure is becoming more frequent and poses significant risks to operational continuity, data integrity, and the stability of the grid. This study presents an integrated approach to cyber-attack detection and data protection in smart grids by leveraging Deep Feedforward Neural Networks (DFNNs) and the NTRUEncrypt lattice cryptosystem. The proposed framework employs DFNNs to identify and classify cyber threats by analyzing system anomalies and network traffic patterns. The detection model is trained and validated utilizing the simulated IEEE bus system in MATLAB Simulink, providing a realistic and scalable environment for benchmarking its performance. To safeguard sensitive communication between grid components, the NTRUEncrypt algorithm is implemented as a quantum-resistant encryption mechanism, ensuring data confidentiality and integrity. Performance evaluation highlights the framework's ability to detect cyber-attacks with high accuracy and low latency while maintaining secure and efficient data exchanges. Metrics such as detection accuracy, false positive rate, encryption speed, and computational overhead are benchmarked to validate its effectiveness. The results demonstrate that the combination of DFNNs for threat detection and NTRUEncrypt for data protection enhances the smart grid's perseverance in the face of cyber threats, establishing a foundation for security, reliable, and future-proof energy systems.