Algebraic attacks are effective for extracting secret keys hidden in white-box procedures. However, existing methods often suffer from limited applicability or excessive complexity. We observed that when randomly generated linear encodings lack full rank, plaintext bytes can be partitioned into multiple equivalence classes. Leveraging this observation and the concept of integral analysis, we strip away the encodings’ protection and construct distinguishers to retrieve keys. Experimental results demonstrate that our attack against the DIBO structure and its variants approaches a success rate of \(100\%\) . Moreover, its time complexity is below \(O(2^{24})\) . Compared to other attack methods, it boasts the lowest time complexity while maintaining a comparable success rate.