How to prevent cheating in online exams?

Original article was published on Artificial Intelligence on Medium

How to prevent cheating in online exams?

5 AI Tools which give cheaters a very hard time.

Photo by Nathan Dumlao on Unsplash

Online learning used to be trendy. Attending university or professional training courses online has become more and more popular. Not only for students but also for employees and lifelong learners. Today, the current global health crisis and the need for social distancing are accelerating this demand. Online teaching and examination are not trendy, they are now becoming mandatory. Universities, corporations and governments are forced to rethink their teaching practice entirely.

At this point, you might think:

“Online learning? Not a big deal.”

“Online courses, presentations and workshops? Doable and already common practice.”

But what about online examination?

It is common kowledge that dishonesty and cheating are a frequent occurrence in any examination. In fact, the word “cheating” in combination with “online exams” is a very frequent search (

So, the key questions are: Can we secure the integrity of online exams without personal supervision? Can we assure a level playing field for all participants of an online exam? And in general, can we prevent cheating?

The answer is yes. Indeed, if we apply the right technology, cheating in online exams can be even harder than in lecturing halls with human proctors. How is that possible? The solution lies in remote and automated proctoring.

Remote proctoring ensures the same integrity for online exams as for offline exams. The three types of remote proctoring are: Live proctoring, record and review and automated (AI enabled) proctoring. In the former two types, a person reviews the footage. Automated proctoring uses AI-based algorithms without the need for any human observation.

Automated proctoring uses smart algorithms which gather and analyse various sources of data. Deep Learning thereby enables the classification and the tracing of illegitimate behaviour like impersonation or the use of unauthorized resources. Thus, the accuracy and the quality of the autonomous system depends on the amount and variety of data points as well as built-in redundancies.

5 AI Tools which give cheaters a very hard time

If you combine a wide range of data tools, any cheating activity in online exams can be detected and prevented. The necessary tools are available on the market, but automated proctoring is not yet widely used. The most crucial tools to enable a level playing field in online exams are:

1) Facial recognition

2) Speech & type recognition

3) Eye & skeleton tracking

4) Browser & app restrictions

5) Plagiarism validation

Facial recognition compares a biometrical model of the candidate’s facial features to a picture of the candidate’s ID document. Thereby the identity of the test taker can be verified. Continuous checks during the test ensure that it is not the candidate’s expert friend who sits in front of the computer taking the test.

Speech and type recognition analyse the voice and keystroke dynamics. It is complementary to the identity verification. The tools secure that the right person writes the answers without someone whispering the solution in the background.

Eye tracking algorithms follow the candidate’s eye movements. They show if the candidate reads something placed next to the screen. Skeleton tracking software can track the candidate’s movement and detects suspicious behaviour.

Now you might think:

“Alright, that sounds all plausible. But how to prevent candidates from simply googling the answer?”

There are browser and app restrictions, which are crucial to prevent the use of unauthorized material. Common options are: Forced full screen mode, clearing of the cache and disabling the copy/paste function.

Still not convinced? In case the candidate still manages to cheat, plagiarism software can do the rest and verify the originality of text.

This all may sound like total surveillance. Indeed, for the time of examination it is. Thus, the beginning and end of the proctoring procedure must be acknowledged and confirmed by the test taker. The examinee must have the supreme authority to stop the remote proctoring — and thereby also the test. As a basic precondition, institutions which handle this sensitive and personal data must follow data protection regulations. Last and most importantly, privacy by design must be the major principle of automated proctoring software. Privacy by design ensures the automated deletion of all personal data after the test.

We can not stop AI from finding its way into more and more areas of our life. But we can adapt to it and use it to enable trust in a digitized world.