SnS2-Based Memristor Arrays for Neuromorphic Computing and Reconfigurable Physical Unclonable Functions

Abstract The increasing demand for energy-efficient and secure on-device artificial intelligence has stimulated considerable interest in multifunctional memory platforms that combine neuromorphic computing and hardware-level security within a unified architecture. Herein, we present a tin disulfide (SnS2)-based memristor array that seamlessly integrates neuromorphic functionalities and reconfigurable physical unclonable functions (PUFs) within a unified hardware platform. The device employs layered SnS2 as the interlayer material, where intrinsic defect states and stochastic conductive-filament evolution govern the resistive switching behavior. Leveraging tunable analog conductance states and robust nonvolatile switching behavior, the memristor faithfully emulates synaptic plasticity and demonstrates high recognition accuracy in neuromorphic computing tasks. Meanwhile, the stochastic formation of conductive pathways during resistive switching generates unpredictable electrical responses, enabling robust and reconfigurable PUF functionality with strong resilience against machine learning attacks. This work provides a promising pathway toward secure neuromorphic hardware for future edge AI applications.