On Coderbyte, we have deployed both industry-standard as well as proprietary methodologies to prevent and detect cheating.
Cheating prevention
To reduce the ability for candidates to cheat, we offer the following capabilities:
- For algorithmic challenges, require candidates to write 100% unique solutions by enabling cheat-proof technology
- For all challenges, mask the challenge title so candidates cannot easily Google for solutions
- Randomize the selection of questions each candidates sees
- You can enable identity verification for any Assessment
- You can enable screenshot-proofing to prevent AI tools from taking screenshots
Cheating detection
If a candidate cheated, we provide several automated and manual detection methodologies. The following are how we display automated cheating detection:
- If a candidate copy and pastes code into the code editor
- If a candidate leaves the code editor tab during an assessment (the length of time impacts our detection algorithm)
- If a candidate plagiarized from other candidates or from ChatGPT
- If a candidate does any suspicious activity, like skipping webcam verifications
Based on a combination of these factors, you will see a summarized cheating detection indicator within any Assessment's dashboard:
- Not detected means that the candidate did not behave in any ways as to indicate that they cheated
- Likely means that the candidate behaved in ways that indicate they likely cheated
- Detected means that the candidate behaved in ways that indicate they definitely cheated
On the individual candidate report, you can see the overview of cheating that was detected:
We also provide the following manual cheating detection methodologies:
- If you have Google search enabled within the code editor, you can see what each candidate Google searched for and clicked by clicking Expand all above the Challenge solutions section
- Watch a video playback of all coding activities to see any suspicious behaviors
- Compare the country of the person taking the Assessment with the expected country of the candidate
- Identify candidates and ask them to explain answers over webcam video
- Enable webcam verification and proctoring to ensure that (i) candidates are who they say they are and (ii) that candidates worked on their solution independently
