In order to solve the problems that the traditional chaotic detection system is easily affected by the transition zone and the detection system is affected by the noise, an inverse phase transition chaos detection method based on high-order cumulants and Duffing oscillator is proposed. Firstly, the Lyapunov index method is used to calculate the critical threshold γ_d of the detection system, so that the the periodic driving force of the detection system is equal to γ_d. Secondly, the signal to be detected is preprocessed by calculating its high-order cumulant, which can reduce the noise power and obtain the amplitude variation of the harmonic signal. Then, the processed signal to be detected is input to the detection system, and the Lyapunov index is used to get the periodic driving force amplitudes corresponding to the reverse phase change. Finally, the amplitude of the signal to be detected and the detected signal-to-noise ratio are calculated according to the difference of the amplitude of the periodic driving force before and after the inverse phase transition. Simulation results show that the proposed method can be used for the detection of weak sinusoidal signals with a signal-to-noise ratio of -50.97 dB, which has a better detection effect than the traditional Duffing oscillator detection system.