I have taught several introductory CS courses and to be honest, I was not interested in playing policeman and checking whether students are cheating or not. Instead, I have established a two tier system:
- For homeworks that had to be turned in, these were corrected by the students themselves. I did some spot checks to warn those who were cheating and to ensure that the corrections are up to par, but didn't really put much effort into chasing cheaters. The homeworks were primarily a feedback for the students and an opportunity to learn. However, to give them an incentive to actually do them, they could pass the exam orally in advance instead of a practical programming exam if they had 80% of homeworks right. That was strong motivation for many of them, because they perceived the oral exam as easier (even though in reality they had to do much more work over the semester for it).
Now, the purpose of the oral examination was simple - to establish whether the homeworks were actually done by that student or not. In my experience, if someone was cheating, he didn't have a clue whatsoever what the code he has handed in does. At best, he could memorize some superficial stuff and do some hand-waving over it. One or two targeted questions over the details of the assignment has always uncovered this. No need for any computerized code comparison tool (which would be always gamed) or tiresome reviews of the homeworks.
- For the regular exam which was always written, practical programming assignment on a computer in the lab (CS exam on paper?? WTF?), I have allowed the students to bring their own code snippets (e.g. from homeworks), use their books, even internet. This essentially makes all what would usually be considered cheating allowed, lessening the burden on me - I did not have to spy on them whether or not they are cheating.
My reasoning was that the students should demonstrate practical knowledge how to solve problems, not whether or not they have memorized stuff (which is what the exam would be about if the books were forbidden). Now, of course, if the student didn't learn anything, the books will not help - they would spend most of their time searching for information and run out of time.
One disadvantage of this approach is obvious - it puts a bigger onus on the examiner to prepare meaningful exams. Assignments like "Implement quicksort" are useless, because the students can find them ready made online or in the book. On the other hand, I do not think it makes much sense to examine whether or not a person can implement quicksort - it is not a real-world problem. Better give them an assignment where the quicksort needs to be used - the clueless one will not find it online so he cannot readily cheat and the smarter one will see the similarity and solve the assignment without problem.
To conclude, I do not believe in the various software to catch cheaters. Especially not in CS - the students are very smart and will be always able to game it. If the teacher is doing their job, this is not needed.