This paper proposes a method for reliably identifying fault nature by injecting disturbance current signals using an external power source, to address the issues of blind reclosing after inter-phase short circuits in distribution networks, which may lead to secondary impact on the system, and the lack of available effective information in adaptive reclosing after protection tripping. After a short circuit fault occurrence, an external power supply injects a low-frequency disturbance current and steady-state disturbance voltage waveform data are extracted. If it is a temporary fault, the inter-line relationship conforms to a capacitive model. If it is a permanent fault, the fault phases exhibit a transition resistance, deviating from the capacitance model. Therefore, the response characteristics of the injected disturbance signal differ significantly between different fault types, thus forming the identification criterion using the difference. This paper introduces a fault property identification criterion based on the Manhattan distance algorithm and the energy relative entropy algorithm. Additionally, an auxiliary criterion is established to further differentiate between permanent faults and cases where faults disappear during signal injection. This approach overcomes the limitations of active fault identification methods in handling fault randomness. PSCAD/EMTDC simulation results show the proposed method has high sensitivity and reliability.