Answers are grounded in this paper and cite their sources.
Experimental platform demonstrates hardware ternary in-sensor encryption, highlighting robust defense against machine-learning attacks in edge sensing.
A central challenge in secure edge sensing is to achieve high‐complexity data encryption directly at the sensing node with minimal hardware overhead. Here we report an integrated one thin‐film‐transistor‐one threshold‐switching‐memristor (1TFT‐1TS) platform that couples ternary sensing encoding with ternary true random number generation to enable in‐sensor ternary encryption, with the encrypted outputs subsequently reinforced through nonlinear ternary encryption to suppress statistical information leakage. The platform is constructed using indium‐gallium‐zinc‐tin oxide (IGZTO) thin‐film transistors and NbO x ‐based threshold‐switching memristors. By exploiting the gate‐bias‐dependent photoresponse of the TFT, optical sensing information can be directly encoded into ternary states. In parallel, the 1TFT–1TS architecture amplifies subtle cycle‐to‐cycle variations in the high‐resistance state of the TS device, enabling binary true random number generation that can be further extended to ternary operation. Building on these capabilities, we implement weak ternary encryption (W‐TE) directly at the sensor end, followed by nonlinear strong ternary encryption (S‐TE) for further security enhancement. Nonlinear ternary operations markedly improve resilience against machine‐learning‐based attacks, lowering attack accuracy to 5.5%. This work establishes a viable hardware route toward secure ternary encryption systems for edge sensing and Internet‐of‐Things applications, and provides a promising strategy for addressing the critical challenge of data security in resource‐constrained edge devices.
No takes yet. Share an insight, caveat, or question.
Zeng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: