Securing the Data Using an Efficient Machine Learning Technique

More accessible data and the rise of advanced data analysis contribute to using complex models in decision-making across various fields. Nevertheless, protecting people’s privacy is vital. Medical predictions often employ decision trees due to their simplicity; however, they may also be a source of privacy violations. We will apply differential privacy to this end, a mathematical framework that adds random values to the data to provide secure confidentiality while maintaining accuracy. Our novel method Dual Noise Integrated Privacy Preservation (DNIPP) focuses on building decision forests to achieve privacy. DNIPP provides more protection against breaches in deep sections of the tree, thereby reducing noise in final predictions. We combine multiple trees into one forest using a method that considers each tree’s accuracy. Furthermore, we expedite this procedure by employing an iterative approach. Experiments demonstrate that DNIPP outperforms other approaches on real datasets. This means that DNIPP offers a promising approach to reconciling accuracy and privacy during sensitive tasks. In DNIPP, the strategic allocation of privacy budgets provides a beneficial compromise between privacy and utility. DNIPP protects privacy by prioritizing privacy concerns at lower, more vulnerable nodes, resulting in accurate and private decision forests. Furthermore, the selective aggregation technique guarantees the privacy of a forest by combining multiple data points. DNIPP provides a robust structure for decision-making in delicate situations, ensuring the model's effectiveness while safeguarding personal privacy.