Furkejuvvon:
| Váldodahkkit: | , |
|---|---|
| Materiálatiipa: | Recurso digital |
| Giella: | |
| Almmustuhtton: |
Zenodo
2026
|
| Fáttát: | |
| Liŋkkat: | https://doi.org/10.5281/zenodo.18687603 |
| Fáddágilkorat: |
Lasit fáddágilkoriid
Eai fáddágilkorat, Lasit vuosttaš fáddágilkora!
|
Sisdoallologahallan:
- <p><span>The confluence of big data, cloud computing, and stringent privacy regulations has made secure outsourced analytics a critical necessity. Distance-based learning algorithms, such as k-Nearest Neighbors (k-NN) and clustering, are fundamental tools for data exploration but face profound scalability challenges when applied to encrypted data. This article provides a comprehensive analysis of the scalability bottlenecks inherent in performing distance-based learning on encrypted big data. We deconstruct the problem by examining the computational and communication overheads imposed by leading encryption paradigms, including Homomorphic Encryption (HE), Garbled Circuits (GC), and Oblivious RAM (ORAM). The analysis reveals a scalability trilemma where data volume, data dimensionality, and security guarantees compete for resources, leading to impractical latencies and bandwidth consumption. Through a quantitative evaluation, we illustrate the exponential growth in processing time and storage overhead as dataset size increases. We further explore the efficacy of contemporary optimization strategies, such as hybrid cryptographic protocols, dimensionality reduction, and secure hardware enclaves. The article concludes by outlining open research problems and charting potential future directions to bridge the gap between the theoretical promise of privacy-preserving machine learning and its practical, scalable deployment.</span></p>