Cloud computing provides scalability, cost efficiency and flexibility of numerous applications. Nevertheless, there remain security issues because of data integrity, confidentiality and cyber threats. Cryptographic measures such as encryption and hashing are secure, but do not have the dynamic threat assessment capabilities. This paper introduces a security platform, which combines deep learning with cryptographic hashing and blockchain technology to enrich key exchanges, data verification, and detection of tampering in cloud storage systems. The suggested key exchanged model builds network based key agreement by modifying Diffie-Hellman approach. It uses artificial noise models of neural networks to ensure high security in the key parameters against nefarious activities. The system adopts a CNN-based encryption scheme as well as a decryption system that had been specially developed to offer a flexible encryption security to cloud-based data. The blockchain technology is implemented as a part of the model because it introduces cryptographic hash verification of the stored data to the ledger system that cannot be changed further, as once it is verified it cannot be altered by an unauthorized party. The set of combined Merkle trees results in efficient efficiency improvements on verification processes by reducing the number of computations needed with preserving fast scalable integrity verification algorithms. The proposed model is also tested by performance standard and security analysis in order to determine its efficacy. Findings indicate that a 256-bit key provides the best PSNR of 30.9778 dB and SSIM of 0.8741 at 32x 32 resolution, and ensures the highest visual similarity and structural similarity after decryption.