Adaptive Machine Learning Model for Dynamic Field Selection

Machine learning is a method of training predictive models using collected data and algorithms to identify correlations between features. However, it faces limitations in protecting data privacy. To address this challenge, we propose a new trapdoor method for marking data based on field combinations to achieve data privacy. Our approach does not require modifying the model itself; instead, we use field labels to exclude users from the model. We add headers to the original data, containing corresponding field combinations, allowing the model to recognize these headers during training. Thus, when predicting data with marked headers, the model can exclude data not belonging to that field combination. Finally, we conducted several experiments using the MNIST dataset to verify the effectiveness of our method. Results show that our approach is not only effective but also time-saving. In conclusion, we offer a new perspective on achieving data privacy.