This paper proposed an islanding detection approach combining phasor measurement unit (PMU) data of branch current and rate of change of branch current. This new islanding detection approach not only can detect the islanding due to the utility end but also can recognize the islanding inside the distribution system, while the existing islanding detection methods only considered islanding on the utility side or inside the distribution system and mostly overlooked an applicable multi-functional cost-effective islanding detection method due to the inappropriate PMU index threshold determination approach. Therefore, the proposed method introduced the ensemble boosting tree method (EBTM) for the determination of the islanding index threshold. The proposed approach is validated on a real active distribution network (ADN) named “Utility Kerteh” located in the state of Terengganu, Malaysia. The system is comprised of a range of distributed energy resources and it has been tested within both the PowerWorld simulator and MATLAB environments, encompassing 17 distinct real electrical scenarios that reflect 5 varied system conditions. Results showed that the proposed method can identify islanding within 20 ms across all cases. Furthermore, it effectively differentiates between islanding and non-islanding situations. A software-in-the-loop (SIL) test system is developed to assess the real-time performance of the proposed method.