SearcharxivSearch

arXiv subjects

Dae-young Kim

Publications and source records attributed to Dae-young Kim.

2 recordsLinked to original sources

TRUCE: TRUsted Compliance Enforcement Service for Secure Health Data Exchange

Organizations are increasingly sharing large volumes of sensitive Personally Identifiable Information (PII), like health records, with each other to better manage their services. Protecting PII data has become increasingly important in today's digital age, and several regulations have been formulated to ensure the secure exchange and management of sensitive personal data. However, at times some of these regulations are at loggerheads with each other, like the Health Insurance Portability and Accountability Act (HIPAA) and Cures Act; and this adds complexity to the already challenging task of Health Data compliance. As public concern regarding sensitive data breaches grows, finding solutions that streamline compliance processes and enhance individual privacy is crucial. We have developed a novel TRUsted Compliance Enforcement (TRUCE) framework for secure data exchange which aims to automate compliance procedures and enhance trusted data management within organizations. The TRUCE framework reasons over contexts of data exchange and assesses the trust score of users and the veracity of data based on corresponding regulations. This framework, developed using approaches from AI/Knowledge representation and Semantic Web technologies, includes a trust management method that incorporates static ground truth, represented by regulations such as HIPAA, and dynamic ground truth, defined by an organization's policies. In this paper, we present our framework in detail along with the validation against the Health Insurance Portability and Accountability Act (HIPAA) Data Usage Agreement (DUA) on CDC Contact Tracing patient data, up to one million patient records. TRUCE service will streamline compliance efforts and ensure adherence to privacy regulations and can be used by organizations to manage compliance of large velocity data exchange in real time.

cs.CR

mhGPT: A Lightweight Generative Pre-Trained Transformer for Mental Health Text Analysis

This paper introduces mhGPT, a lightweight generative pre-trained transformer trained on mental health-related social media and PubMed articles. Fine-tuned for specific mental health tasks, mhGPT was evaluated under limited hardware constraints and compared with state-of-the-art models like MentaLLaMA and Gemma. Despite having only 1.98 billion parameters and using just 5% of the dataset, mhGPT outperformed larger models and matched the performance of models trained on significantly more data. The key contributions include integrating diverse mental health data, creating a custom tokenizer, and optimizing a smaller architecture for low-resource settings. This research could advance AI-driven mental health care, especially in areas with limited computing power.

cs.CL