When sharing data, the theft of students’ mental health privacy data may pose a threat to students themselves, medical institutions, and even society. How to realize the safe storage of mental health privacy data is a problem that needs to be studied in medical data sharing. This paper aims to study a secure storage method of students’ mental health privacy data based on machine learning. Combining the medium k-means algorithm of machine learning with the k-anonymity model, the k-anonymity clustering algorithm is constructed to realize the anonymization of students’ mental health privacy data and complete the primary safe storage. The key is generated by the neural network algorithm in machine learning, and the anonymous student mental health privacy data is encrypted to complete the deep security storage. The results show that the decryption speed of the research method is the fastest, and the implementation efficiency of the security storage algorithm proposed in this paper is high; After the safe storage process, the information loss of the research method is less, and the integrity of private data can be better guaranteed. After the research method is processed, the entropy value is larger, which indicates that the amount of information contained in the processed private data is smaller, and it is more difficult to be stolen.