Investigation of Mobile Machine Learning Models to Preserve the Effectiveness of User Privacy

Machine Learning (ML) has become one of the dominant technologies in the research world. It is being applied without exception in every field where automation and future predictions are required such as cyber security, computer vision, data science, search engines and various other disciplines. The application of ML in search engines creates a high risk of breaching user's privacy because this involves using data gathered from user's browsing history, purchase transactions, searching videos and queries. The user's information gathered from the search engine queries stored in computers, mobiles, other handheld devices is privately uploaded to a centralised cloud location and is then utilised in designing various ML models. As most ML models use a trained model that requires large datasets, user data gathered this way plays an important role in the development of the ML models. This however creates a significant privacy issue for individuals who may not want to reveal their personal information for ML training yet, prevention of this is beyond their access and control. In this article, we focus on the use of ML in mobile devices and address privacy concerns that can be raised by practising ML in mobile devices. The primary area of study in this research is the comparison of ML on mobile devices with ML on the cloud and figuring out its feasibility of becoming an essential ML for preserving user's privacy. Sequentially, this study first explores the need for using the ML algorithm to address privacy issues. Next, a pre-chosen ML algorithm will be tested on mobile devices and cloud to get the comparison outcome that justifies the adoption of privacy-preserving ML model on mobile devices to preserve the user's privacy.