International Research Journal of Education and Technology

Data poisoning attacks on machine learning model reliability, is to identify poisoned attacks in a data set and support clinicians for their future study. A data set is a collection of related sets of information composed of separate items, which can be processed as a unit by a computer. Healthcare data sets include a vast amount of medical data gathered from various healthcare data sources. The healthcare data sets need to be always kept secured, since they can be used further by the doctors or researchers. The chosen healthcare data sets are first run through an ML algorithm called Bayesian Neural Network, which determines the dataset's accuracy. The data set's accuracy can be used to determine whether the data set is poisoned or not. Following that, the dataset is pre-processed, and then the three algorithms named Random Forest, Support Vector Machine and Logistic Regression are used. The highest accuracy producing algorithm is chosen as the best. The best algorithm is chosen as Support Vector Machine due to its high accuracy and is then used to help doctors in the further study of the patient's health condition.

Keywords - Data poisoning, Bayesian Neural Network, Random Forest, Support Vector Machine, Logistic Regression