Today's world has witnessed the accumulation of data through social media like Twitter and Facebook, which have been growing enormously day by day. Using these social media, users tweet or post on many topics from anyplace in the world through the Internet. Tweets are used to analyze positive and negative messages to drive polarity scores and also to predict future trends. These polarity scores could be extracted from Twitter, even though polarity score data are placed under security. The data can be easily hacked, and changing the score results can lead to lots of issues, such as affecting the worldwide economic status, company brands, and the reputations of firms. To tackle these issues, Daniel Bernstein proposed the implementation of Salsa, which has faster encryption because of the quarter round with better data security. In the present work, a novel algorithm is proposed by modifying the Salsa20/4 to further enhance the security of the polarity scores, which is an important requirement of the present world. The proposed algorithm has two stages. The first stage consists of a secret key and the second stage constructs the diagonal keys in N-1 steps. The proposed algorithm is tested with different file sizes to calculate the encryption time. The results show that the proposed encryption algorithm takes relatively large execution time but provides high data security when compared to that of the existing algorithm of Salsa20/4. © 2020 Elsevier B.V., All rights reserved.