Abstract: Cryptographic algorithm identification holds significant importance in the cryptanalysis field. In mainstream block cipher algorithm identifying schemes, the ciphertext feature extraction is typically based on randomness detection standards. To address the need for identifying short samples and further improve the identification accuracy, an identification scheme for block cipher systems based on RF-Light GBM is proposed in this paper. Firstly, ciphertext features are extracted according to the GM/T 0005-2021 randomness detection standard, and the Random Forest algorithm is utilized to rank and filter the importance of high-dimensional data. Then, feature vectors are used to train a Light GBM model to construct a cryptographic algorithm classifier for identifying. This method fulfills the need for efficiently identifying unknown ciphertexts under short sample conditions. The proposed scheme achieves an overall accuracy enhancement of about 12% in pairwise identification experiments, outperforming existing schemes. Multi-classification experiments show that the accuracy rates exceed 73%, verifying the effectiveness and superiority of the proposed scheme in block cipher algorithms identification, which provides a valuable reference for future cryptographic algorithm identifying under more complex encryption modes.