Abstract:

This work introduces a new framework for secure data processing in distributed environments through the combined use of classical secret sharing, strengthened symmetric encryption, verifiable computation, and adaptive monitoring based on deep learning. In this design, a confidential value is divided into independent components, and each component is separately encrypted and validated to prevent disclosure and manipulation. The verifiable computation layer enables the correctness of results to be confirmed without revealing any underlying information. In addition, an intelligent analysis mechanism based on computational vision models monitors encrypted representations and operational traces to detect abnormal or adversarial behavior during processing. The integration of these mechanisms forms a multilayer system that simultaneously ensures confidentiality, correctness, and adaptive oversight, providing a reliable foundation for cloud platforms, large-scale networks, and data-intensive applications.