Wouldn't the malicious alterations introduced in a scenario like that most likely be exploitable defects that could be explained away as mistakes? If they accumulate too much around certain people that's suspicious of course, but it seems like it would often be difficult to downright prove that someone intentionally broke the security of a rather complex system.