Deep Learning-Based Data Sanitization and Restoration for Cloud Privacy Preservation
Rajkumar Patil, Gottumukkala Hima Bindu · Journal of Organizational Computing and Electronic Commerce · 2025
Although cloud computing is increasing in prominence by allowing outsourced and affordable data processing, it raises severe privacy concerns while transmitting sensitive data to cloud servers. Moreover, sensitive data has significant monetary and reputational worth, and any breach of privacy can result in significant financial and reputational loss. Organizations in the financial, health, criminal, social network, and government sectors have been gathering and processing personal data for profit. However, gathering and sharing individuals’ sensitive and confidential information for data mining result in a breach of data privacy. This paper intends to propose a data sanitization and restoration process via a deep learning-based tuned key to ensure cloud security. The data is sanitized by performing the following phases, including data pre-processing, key generation, and key fine-tuning. Data preprocessing includes the extraction of improved statistical features from sensitive data that preserves the originality of the data. Key generation is a subsequent process that takes place in an optimal way via inducing an optimization algorithm, termed as Self Improved Namib Beetle Optimization (SI-NBO) model, to optimize the randomly generated keys under the consideration of parameters like privacy, hiding failure, and preservation ratio. The optimal key is then fine-tuned via the Deep Belief Network (DBN) model to obtain a fine-tuned key. By XORing sensitive data with the key, the sanitized information is obtained. On the other hand, in the restoration process, original data is restored via the same optimal key, which is generated as per the proposed SI-NBO model.