The growing adoption of Internet of Things (IoT) devices has raised serious security challenges, particularly in ensuring secure communication amid increasingly sophisticated cyber threats. While Quantum Key Distribution (QKD) provides a theoretically unbreakable key exchange mechanism based on quantum principles, real-world deployments still face vulnerabilities. To mitigate these risks, this study introduces a hybrid deep learning approach that merges the spatial feature recognition of Convolutional Neural Networks (CNN) with the sequential data processing of Bidirectional Long Short-Term Memory (BiLSTM) networks. The model is trained and tested on the N-BaIoT dataset, utilizing optimization techniques like adaptive learning rates and early stopping to improve detection efficiency. Results demonstrate that the framework achieves high detection accuracy and robust precision-recall performance, successfully identifying intricate attack signatures. These outcomes highlight the promise of combining quantum-based cryptography with advanced machine learning to enhance IoT security and resilience.