In this attack, the actor research the code to discover where and how to change some critical step of the algorithm (or program), causing an undetected and predicted malfunction to achieve his goals. It may be related to leaking the user's key to a remote location or file, or to make the algorithm vulnerable by bypassing several rounds or critical steps (the encryption will be weak), or even changing large portions of code to perform (or not) a specific function.
Therefore, this attack is critical to learn about the target and see where the malicious artifact should work - it usually precedes a code injection attack.
This type of attack on the OBAKE application and the OBAKE-512 algorithm is mostly impractical.