Securing Logistic Regression Model from Data Poisoning Attack using AES and RSA Encryption Techniques
Bhawana S. Dakhare, Lata L. Ragha · 2023
In today's information-exchange-driven world, data privacy, and confidentiality are of the utmost significance. The purpose of the proposed research work titled "Securing Logistic Regression Model from Data Poisoning Attack Using AES and RSA Encryption Techniques", is to show the use of encryption algorithms to provide privacy to data. As machine learning models have turned out to be more prevalent and are used to process sensitive data, ensuring their security is of utmost importance. Machine learning models are susceptible to a variety of threats, according to research, which jeopardizes the models' security. The proposed framework runs on systems that involve client and server. The server which performs inference over data sent from the client and the machine learning model which has been trained on proprietary data. The machine learning model on the server is protected against data poisoning attacks. The proposed system gives a solution for a Server-Side Request Forgery attack against unauthorized access to the machine model. For the privacy of the predicting breast cancer machine learning model on the server, this system employs the encryption methods such as AES and RSA.