Giant Trevally Optimization Driven CNN Transformer Framework for Intrusion-aware Secure Cloud Storage

Cloud-based systems need intrusion detection mechanisms that are strong enough to protect against unauthorized access to the system and data, requiring cryptographic techniques.This study suggests an Intrusion Aware Secure Cloud Storage and Retrieval Framework that integrates a lightweight CNN-Transformer model using attention refinement for multi-class intrusion detection, quantum convolutional neural network (QCNN)-based dynamic encryption as well as attribute-based encryption (ABE) for extra fine access control namely Giant Trevally Optimization method for joint hyperparameter tuning.The performance of the proposed approach was evaluated using the NSLKDD dataset, which contains about 137,000 network traffic records categorized into five classes (Normal, DoS, Probe, R2L and U2R).For binary classification, it achieved 98.9% accuracy and an area under the curve (AUC) of 0.996, a clear indication that for these data it's able to distinguish well between benign and malicious instances.Using the five-class classification, it attained an overall accuracy of 96.4% and a macroF1 score of 94.3%.Additionally, it showed better detection performance against minority attack classes R2L (91.7%) and U2R (89.4%).The encryption scheme based on QCNN provided ciphertext entropy close to the ideal value (7.999), and it showed avalanche effect with diffusion quality of more than 50% denoting strong diffusion properties.The computational analysis revealed moderate model complexity (2.8M parameters) and real-time inference capability (1.34 ms per sample), validating practical deployment feasibility.The system that was provided enables comprehensive generation and scalability for the management of secure cloud data.This is done by tightly coupling intrusion intelligence with dynamic cryptographic key management.The framework is further validated on modern datasets (UNSW-NB15 and CICIDS2017) to demonstrate generalization capability.