Privacy-Preserving Machine Learning for Mental Health Prediction Using Homomorphic Encryption
Bibliographic record
Abstract
Student mental health issues, such as stress, anxiety, and depression, are increasingly prevalent in academic institutions, significantly affecting well-being and academic performance. Recent machine learning (ML)-based systems have demonstrated promise in predicting mental health conditions using survey data, but these approaches often process sensitive information in plaintext, risking privacy breaches or relying on centralized data storage vulnerable to leaks. Homomorphic encryption (HE) has been proposed for secure ML, but existing implementations either focus on simpler datasets (e.g., numerical/IoT data) or incur impractical computational overhead (e.g., high RAM usage or prolonged training times) for real-world mental health applications. To address these gaps, we introduce a privacy-preserving predictive model for student mental health using logistic regression trained directly on encrypted data via the TenSEAL library. Our work uniquely combines a leveled fully homomorphic encryption (FHE) scheme to ensure end-to-end confidentiality, replacing the standard sigmoid with a quadratic approximation for homomorphic compatibility. We also perform a comprehensive efficiency analysis that evaluates RAM usage and training time across polynomial-modulus degrees to balance security and practicality, a trade-off underexplored in prior HEbased mental health studies. Experimental results show that our encrypted model achieves 84% accuracy (vs. 96% unencrypted) with minimal performance loss, while benchmarks demonstrate scalable resource consumption. This work advances the feasibility of implementing FHE in sensitive domains such as mental health, offering a rigorous template for privacy-preserving ML without compromising predictive utility.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.007 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".