Detection of high-impedance faults (HIFs) is significant for protecting active distribution networks and human safety. However, existing HIF detection methods show low accuracy and reliability, and protection remains a great challenge in active distribution networks. In this article, a novel HIF detection method is proposed by considering nonlinear arc distortion features in active distribution networks. First, according to the dynamic processes of arc quenching and restriking, the changing characteristics of arc current are analyzed theoretically. Then, based on the theoretical analysis of arc distortion features, two distortion indicators are defined. The first is the arc time-periodic indicator, and the second is the nonlinear distortion ratio. Finally, a HIF detection criterion is developed by combining the two distortion indicators. 150 sets of simulation and experimental results validate the reliability and applicability of the proposed detection method. Across all tested scenarios, the proposed method achieves a detection rate exceeding 98% and maintains a false alarm rate below 2%. In addition, the proposed method can accurately differentiate between HIFs and normal transient disturbance events such as load switching, capacitor switching, and transformer inrush current, ensuring the security of the proposed method.