Computationally Efficient Approaches for Privacy and Security of Big Data

Abstract Big data demands the cloud for storage, but organizations are not relying on the cloud, because of security and privacy reasons. When choosing the cloud for data storage, users worry about issues such as leakage of personal information, unauthorized user access and modification of data, and malicious behavior in cloud data access. Privacy leaks should be avoided during data analysis. Data integrity must be maintained to avoid data modification. Also need machine learning models to detect malicious activity patterns from users in the cloud to ensure data security. Therefore, big data requires technology that protects privacy and security with minimal investment of time and space. Hence approaches using anonymization, sanitization, integrity, and machine learning techniques are proposed. Anonymization can protect personal data from being disclosed upon receiving the data. The use of map-reduce along with anonymization reduces the computational overhead. Sanitization of type encryption/decryption with weighted fuzzy c-means clustering(WFCM) can protect privacy and security with less computation time. Elliptic curve with Diffie–Hellman (ECDH) algorithm is used to protect the integrity of the data with better accuracy than the other cryptosystem algorithms. To protect and safeguard the transactions in cloud applications, a novel solution using Machine Learning (ML) approach with Convolution Neural Networks and a support vector machine is employed. The efficiency of the classifier ability is measured using parameters such as precision, recall, and F-score. Therefore, all these techniques allow this research to protect the privacy and security of big data.