Using Cryptography Methods Against Text Classification
Articles > Article · 2024
Authors: Ahmet Emre Ergün, Özgü CAN
International Journal of Multidisciplinary Studies and Innovative Technologies, Refereed Journals of Other Institutions, International, Diğer endeksler, 2024, vol 8, pp 92-98
Summary
<p>This article explores how cryptographic techniques can be used to ensure the confidentiality of data against machine learning classification algorithms. The study focuses on the Mental Health Corpus dataset, which contains text and tag columns related to mental health issues. Random Forest (RF), Decision Tree (DT) and Support Vector Machine (SVM) classification algorithms were used to classify the texts. To reduce classification accuracy, character shift, which is a cryptography method, is applied. Results show that character shifts greatly reduce classifier accuracy, with shifts as small as 1 character reducing accuracy by more than 30% across all models. The findings demonstrate the potential of cryptographic methods to increase data confidentiality and security in a variety of areas, especially where sensitive information is involved.</p>