Research on Data Processing and Algorithm of Personalized Learning from the Perspective of Privacy Protection

With the advent of the digital age, personalized learning has gradually become a hot topic in the field of education. In the process of personalized learning, a large number of learning data processing is involved, however, there is a huge contradiction between data value mining and data privacy protection, which seriously restricts the release of data value. In view of this contradiction, some teams have proposed a data analysis architecture that separates the debugging environment from the operating environment, and realizes the security concept of “data is not moving, the program is moving, and the data is available and invisible”. The realization of data value mining under the premise of protecting data privacy has been widely used in government affairs, medical care and other industries. Under this privacy protection architecture, because the data and training process are inaccessible, it is difficult to debug the deep neural network model to improve its effect when the effect of the model does not reach the expected goal. Existing debugging methods require manual collection and analysis of metadata generated during model training to assist data analysts in debugging models. These methods cannot be directly applied to the privacy protection scenario, because the metadata generated by the manual collection and analysis model will cause privacy disclosure. From a deep learning perspective, this paper explores how to effectively process learning data in the context of personalized learning while ensuring that the privacy of learners is adequately protected. We will first analyze the privacy challenges existing in personalized learning data processing, then introduce the data processing methods based on deep learning, and solve the problems existing in the model and improve the model effect under the premise of invisible data through the neural network model debugging technology under the framework of separation of debugging environment and operating environment. Finally, the reliability of the model is proved through experiments.