For the promotion of submarine′s magnetic silencing ability, it is necessary to monitor the submarine′s permanent magnetic field immediately, and a reflection method of submarine′s internal and external magnetic field based on LS-SVM(least squares support vector machine) was proposed. Combined with internal and external reflection method and LS-SVM theory, an inside-out reflection model of submarine′s magnetic field was established by optimizing the model parameter with CV (cross validation). With the variation in the vertical component of the submarine′s external permanent magnetic field as an object of analysis, the extrapolation answers of simulation and hull experiment agreed well with the standard value. Compared to the RBFNN (radius basis function neural network), the proposed method has better generalization ability and extrapolation accuracy apparently, fits more in engineering facts, and can provide useful guidance in the research for closed-loop degaussing technology.