Social media platforms like Facebook and Twitter generate large amounts of data daily. To analyze sentiment (neutral, negative, positive) and create polarity scores, secure data handling is critical, as data breaches can affect financial and product firms. To address this, D.J. Bernstein proposed the Salsa family of encryption algorithms with faster processes. This paper introduces a novel methodology that modifies Salsa20/4 to improve data security, consisting of three steps: column operation of the matrix, multiplication with a top-secret key, and shifting diagonal elements. The proposed methodology provides enhanced data security compared to Salsa20/4, though it requires more computation time. © 2020 Elsevier B.V., All rights reserved.