Comparative Evaluation of Privacy-Preserving Association Rule Mining Algorithms for Distributed Healthcare Data

Privacy-preserving data mining has become increasingly important in distributed healthcare systems, where multiple institutions must collaborate to extract meaningful knowledge while protecting sensitive patient information. Association Rule Mining (ARM) is widely employed for discovering hidden relationships within medical datasets; however, conventional mining approaches often require data sharing that may compromise patient privacy. Numerous Privacy-Preserving Association Rule Mining (PPARM) algorithms have been proposed to address this challenge, yet their comparative performance under a common evaluation framework remains insufficiently explored. This paper presents a comparative evaluation of three representative PPARM algorithms: ElGamal Homomorphic Encryption-Based PPARM, Distributed Paillier-Based PPARM, and the PPARM framework incorporating the Transaction Hewer Allocator (THA), Facile Hash Algorithm (FHA), and ATid-based local mining. The algorithms were evaluated using Chronic Kidney Disease and Heart Disease datasets in a simulated distributed healthcare environment. Their performance was analyzed with respect to privacy preservation, computational efficiency, communication overhead, scalability, deployment complexity, fault tolerance, and overall suitability for multi-cloud healthcare systems. The comparative analysis shows that while the ElGamal and Distributed Paillier-based approaches provide strong cryptographic privacy guarantees, they incur higher computational and communication costs. In contrast, the THA+FHA+ATid framework achieves a more balanced trade-off between privacy, efficiency, scalability, and practical deployment through lightweight transaction fragmentation, deterministic hashing, and distributed local mining. The findings demonstrate that the selection of an appropriate PPARM algorithm should depend on the operational requirements of the healthcare environment, and they provide practical guidance for researchers and practitioners in selecting privacy-preserving association rule mining techniques for secure distributed healthcare data analysis.