You run a monthly exam for 400 students online. By the time the window closes, a screenshot of question 7 is circulating in a WhatsApp group, and a third of the class scored above 95%.
Most teachers react by asking for tighter surveillance. Camera monitoring, blocked tabs, "the system should detect it". Some of that helps. A lot of it doesn't, and some of it punishes honest students more than cheaters.
This guide is realistic about what a browser-based exam can and can't do, and shows the measures that actually move the needle for a teacher or education centre in Egypt and the Gulf.
What a browser-based exam can and can't do
Start with the honest boundaries. Your exam page runs inside a browser on a device you don't control. It sees what happens in that page, and very little else.
| Measure | What it really does | What gets around it |
|---|---|---|
| Server-side timer | Time keeps running even if the page is refreshed or closed | Nothing, as long as time is tracked on the server |
| Random questions from a bank | Each student gets a different set | Groups pooling answers across many students |
| Shuffled question and option order | "The answer to 5 is B" becomes useless | Sharing the answer text instead of the letter |
| Tab and window focus logging | Records when the student left the exam page | A second device, which it can't see |
| Full-screen request | Asks for full screen and logs when it's exited | Phones where support is limited, second devices |
| Blocking copy, paste and right-click | Small friction inside the page | Screenshots, phone photos, typing the question |
| One attempt per account | Stops retakes to fish for answers | A second paid account |
Notice what's missing: nothing here can see a second phone, a notebook or a helper in the room. Screenshots can't be reliably blocked in a browser either. Anyone promising otherwise for a normal web exam is overselling.
Design beats surveillance: build a real question bank
The single most effective anti-cheating tool is a question bank big enough that sharing answers stops being useful.
- Size. A practical target is three to five questions in the bank for every question on the exam. A 20-question exam draws from 60 to 100, tagged by chapter and difficulty.
- Balanced random draws. Each student gets, say, 8 easy, 8 medium and 4 hard questions from the right chapters. Different questions, same difficulty.
- Shuffling. Randomise both question order and answer order.
- Parametric questions. For maths, physics and chemistry, generate numbers per student. The method is identical, the answer isn't, so a shared answer is simply wrong.
- Formats that resist sharing. Ordering, matching, numeric entry and short written answers are harder to pass around than a single letter.
- Open-book thinking. Assume students have their notes open. Ask them to apply ideas, not recall definitions.
Timing and access rules that close common loopholes
Many leaks happen through timing, not technology. A few rules fix most of them:
- A fixed window. The exam opens at 8:00 pm and closes at 8:40 pm. Late starters get less time, not a fresh 40 minutes.
- Time tracked on the server. Refreshing the page or switching devices doesn't reset the clock.
- Results after the window closes. If the first finisher sees correct answers at 8:15, the rest of the class will too.
- One question per screen for short quizzes. Optionally with no going back. Students dislike this on long exams, so use it where it fits.
- Reasonable time per question. Tight enough that looking everything up costs marks, loose enough that slow readers aren't punished.
Browser signals: log them, don't overrate them
A platform can record when the exam tab loses focus, when full screen is exited and when the page is reloaded. That data is useful. It is also noisy.
A notification popping up, a weak connection or a student's phone ringing can all trigger "left the page". On phones, full-screen support for web pages is inconsistent, so you can't rely on it at all for students testing on mobile.
A fair policy: warn on the first event, flag the attempt after repeated events, and let a person review flagged attempts. Don't auto-fail on the first tab switch.
The stronger signals often come from the answers themselves:
- A student answering 20 hard questions in four minutes with full marks.
- Two students with the same unusual wrong answers in the same order.
- A sudden jump from 40% to 98% with no change in homework.
None of these prove cheating. They tell you whom to talk to.
Camera proctoring and lockdown browsers
Beyond the browser, there are heavier tools. Proctoring services record the webcam or take snapshots and flag suspicious behaviour. Lockdown browsers are separate apps that restrict what the device can do during the exam.
They have a place, mainly for certificates and professional courses where a score carries real weight. For secondary students they bring real costs:
- Recurring fees, usually per exam or per student.
- A webcam and a stable connection, which many students on mobile data don't have.
- False flags that someone has to review.
- Privacy. Recording minors at home needs clear parental consent and careful storage.
- Still no protection against a second phone held below the camera.
For most teachers, better questions and a short oral check achieve more.
Match the setup to the stakes
| Exam type | Stakes | Sensible setup |
|---|---|---|
| Quiz after each lesson | Low | Small bank, shuffling, instant feedback. Don't police it. |
| Weekly or monthly exam | Medium | Fixed window, larger bank, results after closing, focus logging |
| Term exam or ranking | High | Supervised: in the centre on devices or on paper |
| Paid certificate | High | Proctoring or a live oral check, plus a strong bank |
If you run an education centre, the high-stakes exams can run in your hall on tablets or a computer lab, using the same question bank as the online ones. We cover that hybrid setup in running an education centre online and offline in one system.
Common mistakes
- Giving everyone the same 20 questions in the same order.
- Showing correct answers the moment a student submits, while the exam is still open.
- Running the timer only in the browser, so a refresh resets it.
- Auto-failing on the first tab switch and punishing students with poor connections.
- Treating online scores as equal to supervised ones for prizes and rankings.
- Announcing that "the system detects all cheating". Students will test it, and they will find out it doesn't.
Questions people ask
Can the platform tell if a student opens another tab?
It can log that the exam page lost focus, with the time. It can't stop it, and it can't see a second device. Treat the log as a signal for review.
Can screenshots be blocked in an online exam?
Not reliably in a browser. Assume questions can be captured, which is why random banks and results released after closing matter more than blocking.
Is AI camera proctoring worth it for school students?
Usually not. It adds cost, needs good connections and raises privacy concerns with minors, and a second phone still defeats it. It fits certificates and professional exams better.
How big should my question bank be?
A practical target is three to five bank questions for every exam question, spread across chapters and difficulty levels. Grow it gradually after each lesson.
Can the same exams run inside my centre too?
Yes. The same bank can serve online exams and supervised exams in your hall on tablets or a lab, so marks land in one student record.
Where to go from here
Pick your next exam and change two things: draw it from a bank, and release results only after it closes. Those two steps alone cut the most common leaks. If you're planning a full platform, the complete guide to a teacher's e-learning platform and our article on protecting course videos cover the other half of the picture.
Exams and question banks are a core part of the education platforms Unilira builds. If you want to talk through how your exams run today and what to change, book a meeting.
