Currently the researchers use side channel detection method based on dimension reduction to detect the hardware Trojan, which will lost the critical characteristics information of hardware Trojan after filtering or PCA, and caused a huge computation in the subsequent modeling operations. Differs from this traditional detection method, this paper presents a hardware Trojan detection technology based on extreme learning machine (ELM), it can fully retain the useful information and the template is established intelligently by the neural network to avoid the artificial modeling inaccurate. Finally, using the self-developed FPGA experiment platform we collected the side channel current information of the target chip, and analyzed the data by MATLAB. Results show when detecting the hardware Trojan only occupied 0.15% resources, the success rate can reach about 90%.