AbstractReal‐time intrusion detection system based on the von Neumann architecture struggle to balance low power consumption and high computing speed. In this work, a strategy for network intrusion detection system based on the WO3–x/WO3–x‐Ag/WO3–x structured optoelectronic memristor overcoming the aforementioned issues is proposed and demonstrated. Through the modulation of electrical signals, the memristor successfully simulates a series of important synaptic functionalities including short‐term/long‐term synaptic plasticity. Meanwhile, when subjected to light stimulus, it demonstrates remarkable synaptic behaviors in terms of long/short‐term memory and “learning‐forgetting‐relearning.” Based on this memristor array, a convolutional neural network is constructed to recognize abnormal network records within the KDDCup‐99 dataset accurately and efficiently. The power consumption (10–6 W) is over seven orders of magnitude lower than that of central processing unit, etc. Subsequently, an intrusion detection system is established to integrate collection, processing, and detection of real‐time network data, successfully classifying various types of network records. Hence, this work is expected to promote the development of high‐density storage and neuromorphic computing technology, and provides an application idea for intelligent electronic devices.