Temperature monitoring underpins the protection and condition assessment of distribution transformers, whose dominant failure mode is overheating-induced insulation aging; this makes temperature data the prime target of false data injection (FDI) attacks. Because FDI acts on data rather than on the physical asset, this paper exploits the winding structure, which data injection cannot alter, as a physical trust anchor: a structural fingerprint is extracted from port measurements through frequency-dependent harmonic resistance; the systematic temperature-response differences between wire-wound and foil-wound windings build an independent physical cross-check channel; and a lightweight PBFT consortium blockchain notarizes the fingerprint baseline so that a compromised local database cannot rewrite it. Structural, physical, and command consistency indicators classify input-, output-, and model-parameter-tampering attacks. Finite-element simulations and 100-round randomized attack tests show a 9.6% structural temperature deviation that provides a clear physical screening margin: the three attack types are detected and classified within 4 s at a 100% detection rate (95% CI [96.4, 100]%) and a 0.3% false-alarm rate, while a stealthy ramp injection is caught after about 105 min and a sub-threshold constant offset is caught by trend monitoring in about 4 h. A seven-node chain implemented with real ECDSA-secp256k1 signatures achieves 0.42–0.90 s finality per notarization, 8.9–19.0 notarizations/s, and a 0.47 s view change, two orders of magnitude above the field notarization demand, meeting lightweight deployment requirements.
Multimodal clustering achieves outstanding performance in various applications by aggregating information from heterogeneous devices. However, previous methods rely on strong-notion distances to fuse crossmodal complementary knowledge, established on a fragile assumption about the existence of a ubiquitous non-negligible intersection between heterogeneous manifolds of modalities. Due to this unstable theoretical basis, previous methods are essentially challenged by limited performance on general multimodal data. To address this challenge, an optimal-transport-based multimodal clustering (OTMC) method is defined as the optimal transport (OT) from multimodal data distributions to clustering distributions, which leverages a weak-topology measure to capture complementary knowledge with clear discriminative structures. OTMC consists of a modality-specific OT delivering private structures and a modality-common OT delivering shared structures, which transports category structures scattered in manifolds of each modality and all modalities to common prototypes, respectively. Furthermore, variational solutions to OTMC are derived by matching the data-prototype joint distribution, which induces the multimodal OT clustering network, to capture discriminative structures. Finally, the experimental results from four real-world datasets demonstrate the superiority of OTMC, helped by never relying on the phantom of heterogeneous manifold intersections. In particular, OTMC obtains 92.15% ACC, 84.96% NMI, and 83.35% ARI on Handwritten, improving by 2.25%, 2.82%, and 3.28%, respectively.