JMIR Publications

ABSTRACT

Background:

Health data such as medical imaging and associated reports is routinely transferred between hospitals, laboratories, contract research organizations, and biotechnology companies. These transfers often cross intermediary environments—public networks, third-party infrastructure, portable media, and cloud storage—that neither endpoint fully controls. Conventional secure transfer relies on a single protected channel (for example, Transport Layer Security), whose guarantee ends when the connection closes: it does not survive loss of the route, does not prevent an intermediary from observing or withholding the full payload, and leaves no verifiable evidence after delivery. For regulated health data, a transfer must be not only confidential but also recoverable, attributable, and auditable.

Objective:

The objective of this study was to design and evaluate a transfer model that preserves confidentiality, integrity, availability, and regulatory-grade evidence when regulated health datasets move between two controlled environments across partially trusted intermediaries, and to make the evaluation openly reproducible.

Methods:

We designed the Biotech Health Data Resilient Replication Model (BHDAM-R), which replaces the single trusted channel with a set of independent channels and moves the trust boundary from the transport to the data object. A dataset is sealed with authenticated encryption under a recipient-bound key, dispersed into k-of-n erasure-coded shards carried over heterogeneous channels (networks, object stores, secure copy, and physical media), and reconstructed only in the recipient environment after cryptographic verification. An optional all-or-nothing transform raises the reconstruction threshold to a confidentiality boundary. Each transfer is bound to a sender-signed manifest and produces a recipient-signed delivery receipt and transfer trust states (R0–R5). We implemented an open reference implementation and evaluated it through an automated adversarial and fault-injection test suite, Monte Carlo recovery simulation validated against an exact binomial model, per-stage and throughput measurements, and comparison against a TLS-style authenticated-encryption baseline and a classical dispersed-storage baseline. We additionally performed a multi-transport deployment validation over real cloud object stores, secure copy, and a physical channel.

Results:

All 14 functional and fault-injection tests were detected or rejected, including reconstruction below the k-of-n threshold, tampered and replayed shards, wrong-key rejection, whole-transfer replay refusal, and signed-receipt verification. Measured recovery under simulated channel failure matched the binomial availability model (maximum absolute deviation 0.036 at 800 trials per point); for example, a 4-of-6 profile recovered with probability 0.984 at a 10% per-channel failure rate versus 0.900 for a single channel. Sealing and dispersal sustained on the order of 90 to 230 MB/s across the measured payload sizes (1–100 MB); the directed-transfer assurance layer added approximately 60% to 82% over a bare dispersed-storage core, and storage overhead followed n/k exactly. A single transfer of a synthetic imaging dataset was reconstructed byte-for-byte from 10 of 15 shards distributed across five transport technologies and two public cloud providers after deliberate loss of five shards. All result tables, figures, and the signed evidence package are reproducible from the published artefact.

Conclusions:

BHDAM-R converts health data movement from a single fragile transfer into a multi-route, evidence-producing process that preserves confidentiality, integrity, availability, and verifiable provenance without relying on the security of any single intermediary environment. The model composes established cryptographic primitives into a directed, auditable transfer mechanism suited to regulated health-data exchange. Evaluation was performed with synthetic datasets and simulated or controlled channels; validation in a live clinical workflow, bounded-memory streaming for multi-gigabyte payloads, and persistent cross-session replay protection are identified as future work.