This paper aims to realize the simple and fast nondestructive inspection of drug packaging materials in judicial identification. 40 groups of drug packaging boxes from different manufacturers and different places were tested by X-ray fluorescence spectrometry (XRF) under the conditions of 50 kV voltage, 30 μA current and 1.5 kW power with Rh as the anode target. Based on the chemical composition of pharmaceutical cartons, the samples are labeled and a Monte Carlo Algorithm (MC) optimized Support Vector Machine (SVM) classification model was established to simulate the penalty factors for optimization. At the same time, the divide-and-conquer algorithm was combined to realize the half search, so that the iterative process had the ability of self-learning. Finally, the penalty factor group with both fitting and derivability was obtained based on K-fold cross-validation. The computer simulation results show that when the three SVM penalty factors are set to 933, 280 and 732, the MC-SVM model can achieve the fitting of 100% of the training sets and the classification of 90% of the prediction sets, and the lowest Loss value of the Hinge Loss function is 0.0938. This method can provide some reference for the inspection of drug packing box material evidence, and it also provides a new idea for parameter optimization of SVM.