Software engineering education becomes popular due to the rapid development of the software industry. In order to reduce learning costs and improve learning efficiency, some online practice platforms have emerged. This paper proposes a novel test code plagiarism detection technology, namely MAF, by introducing bidirectional static slicing to anchor methods under test and extract fragments of test codes. Combined with similarity measures, MAF can achieve effective plagiarism detection by avoiding a large amount of unrelated noisy test codes. The experiment is conducted on the dataset of Mooctest, which so far has supported hundreds of test activities around the world in the past 3 years. The experimental results show that MAF can effectively improve the performance (precision, recall and F1-measure) of similarity measures for test code plagiarism detection. We believe that MAF can further expand and promote software testing education, and it can also be extended to use in test recommendation, test reuse and other engineering applications.