A Fully Homomorphic Encryption based approach for Privacy Preserved Pre-processing of Medical Transcripts.

Natural Language Processing holds immense potential for extracting insights from healthcare data, but it demands stringent privacy protection. Privacy-preserving NLP techniques, notably Fully Homomorphic Encryption (FHE), offer a path to advanced analytics while preserving patient data confidentiality. This paper marks the culmination of our initial phase, focusing on data preparation and encryption. We’ve employed FHE with the CKKS scheme to ensure data remains encrypted. Performance evaluation adapts multiclass classification metrics, addressing the distinct nature of healthcare data. As we conclude this phase, we emphasize that this is only the beginning. Subsequent steps will focus on developing NLP models trained on encrypted data. Our work highlights the intersection of data privacy and advanced analytics in healthcare, ultimately benefiting healthcare providers and patients. [ABSTRACT FROM AUTHOR]