The rapid digitalization of healthcare is driving medical data sharing across hospitals, departments, regional centers, and health authorities. In such cross-domain settings, secure sharing requires fine-grained authorization, chain-confirmed revocation, auditable access, and accountability for content leakage after legitimate use. Existing access-control schemes often fail to capture real healthcare administrative hierarchies, impose heavy computation on lightweight clinical terminals, or stop at key-level accountability. To address these challenges, we propose H-Trace, a hierarchical access-control and traceability framework for medical data sharing. H-Trace aligns cryptographic delegation with healthcare administrative structures through Hierarchical Ciphertext-Policy Attribute-Based Encryption (H-CP-ABE), and decouples policy enforcement from bulk data encryption using a Key Encapsulation Mechanism–Data Encapsulation Mechanism hybrid architecture. Gateway-assisted transformation shifts pairing-intensive computation away from resource-constrained terminals, while a permissioned consortium blockchain maintains revocation states, audit records, session metadata, and watermark anchors. For ex-post leakage investigation, H-Trace embeds session-bound invisible watermarks into released medical images, binding the released content to user identity, access session, and on-chain evidence. Security analysis shows that H-Trace achieves selective IND-CPA security and resists collusion attacks and out-of-scope delegation. Prototype evaluations demonstrate sub-millisecond post-transformation secret-recovery overhead in the tested setting, efficient protection of large medical files, and recoverable watermark tracing under typical distortions while maintaining diagnostic readability.
Wanbin Liu, Ye Lu, Wenyi Chang et al.· Journal of King Saud Univers...· 0 citations
End-to-end speech language models increasingly represent user speech with speech tokens rather than relying exclusively on cascaded ASR--LLM--TTS pipelines. Although these tokens support expressive and low-latency spoken interaction, they may also preserve sensitive speaker characteristics. We investigate whether exposed speech tokens leak voiceprints and formulate this risk as a speaker inversion attack. We introduce Audio BERT (AuB), a trainable model that constructs token embeddings from discrete codebooks and aggregates them into speaker-sensitive representations, and propose SpInv, a two-stage inversion method built on AuB to recover embeddings in the space of an attacker-specified speaker encoder. We evaluate Moshi, Higgs3, Kimi-Audio, and Qwen3-Omni using speaker-disjoint protocols on the VoxCeleb dataset. Extensive experiments show that, with only three seconds of frontend output, SpInv achieves cosine similarities above 0.70 in the attacker-specified speaker-encoder space.
Ye Lu, Yihan Yan, Zhaoyang Zhang et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.