Protecting wireless network nodes from intrusion is crucial for maintaining overall network security. Many current intrusion detection methods rely on advanced machine learning algorithms but may suffer from overfitting, leading to reduced accuracy in detecting intrusions. To address this issue, a unique approach has been proposed that utilizes isolated random forest for intrusion detection in wireless network nodes. This method involves extracting feature data from wireless network call and state information, followed by dimensionality reduction. A time-stamped Markov model is then used for network segmentation and local feature coding of intrusion data. Subsequently, the Gaussian mixture model clustering algorithm is employed to partition the data into different clusters, each trained with isolated random forest classifiers. Experimental results have shown the efficacy of this method in accurately detecting intrusions in wireless network nodes, demonstrating its viability and dependability.