Text Classification using Machine Learning Algorithms on Encrypted Data

Text classification is a fundamental activity in natural language processing (NLP), that includes classifying text resources into predetermined groups. With the growing concerns over data privacy and confidentiality, there is a need to perform text classification on encrypted data to protect sensitive information. In recent years, homomorphic encryption has emerged as a powerful technique that makes computations possible on encrypted data without compromising its confidentiality. This paper provides an overview of text classification using machine learning algorithms on encrypted data. In this research work, we have taken a data set of movie reviews. The data set is cleaned by removing stop words and performing stemming, lemmatization, and tokenization. We utilized a semantic model with inverse document frequency weighting and term frequency weighting to represent the data (TF-IDF). For the classification of the data set, we have applied six machine learning algorithms Logistic Regression, Naïve Bayes, KNN, Support Vector Machine, Random Forest, and Decision Tree. The plain text is encrypted using a somewhat homomorphic encryption technique called the RSA method. The ciphertext is subjected to the machine learning algorithms mentioned above. We have compared the results for accuracy. The Random Forest algorithm was found to give 75.56% accuracy compared to other algorithms on encrypted data. In this work, we have calculated the time taken by each algorithm for execution on plain text and ciphertext. It was observed that the Random Forest method took more time and the KNN method took less time comparatively.