Every malpractice technique in online exams has a technical counter. Institutions that run credible online exams don't rely on trust — they deploy the counters in layers, so each cheating path hits a control. Here is the map.
The malpractice-to-counter map
| Malpractice | Counter |
|---|---|
| Impersonation / proxy candidate | Photo ID + facial match at entry, continuous face recognition |
| Person helping off-camera | AI multiple-face detection, 360° dual-camera room coverage |
| Phone, notes, smartwatch | AI object detection with flagged snapshots |
| Googling in another tab | Secure browser — tab switching blocked and logged |
| Copy-paste of questions/answers | Clipboard blocked by the secure browser |
| Answer sharing between candidates | Multi-set papers, randomized question order |
| Two sessions on one login | Concurrent-login prevention |
| Post-exam denial ("I never did that") | Timestamped recordings and audit logs |
The layers, briefly
Lock the machine
A secure browser turns the candidate's device into an exam terminal: one locked window, no tab switching, no copy-paste, no screenshots, unauthorized app launches flagged. Second screens are detected.
Verify the person
Identity is checked at entry (photo ID capture, facial matching, proctor approval for high-stakes exams) and continuously during the session, so the person who started is the person who finishes. Center-based exams add fingerprint + photo ID + hall ticket verification at the gate.
Watch the session
AI proctoring analyzes the feed continuously — faces, objects, audio, window switches — and auto-flags incidents with snapshots. Live proctors can chat, pause or terminate. 360° dual-camera mode uses the candidate's phone as a second camera after a pre-exam room scan, covering the desk and room a webcam can't see. It all works down to 0.1 Mbps using image modes.
❌ What a single webcam sees
- The candidate's face, straight on
- Nothing on the desk below frame
- Nobody standing beside the screen
- No second device out of view
✅ What layered monitoring sees
- Room scan before the exam starts
- Desk and surroundings via second camera
- Every extra face and device, auto-flagged
- Every tab switch and app launch, logged
Devalue collusion
Randomization makes sharing pointless: multi-set generation (unique pools or shuffled order) means adjacent candidates hold different papers, and per-question timers close the lookup window.
Keep the evidence
Every flag lands in a proctoring report — face counts, objects detected, switch counts, per-snapshot timestamps — and sessions record for audit. When a result is disputed, the institution answers with evidence. Deeper logging patterns are covered in our security techniques guide.
Proctored online exams run inside the ePravesh admission pipeline — application, hall ticket, exam, merit list — with descriptive answers flowing to onscreen evaluation. Book a free demo to see the counters working in a live exam.
Frequently Asked Questions
What malpractices happen in online exams?
Impersonation, help from people off-camera, phones and notes, tab-switching to search engines, answer sharing between candidates, and running two sessions with one login — each needs a specific technical counter.
How does AI detect cheating in online exams?
AI monitors the video feed continuously for multiple faces, missing faces, phones and other objects, tracks tab switches and app launches, and auto-flags each incident with a timestamped snapshot for review.
Can online exams prevent impersonation?
Yes — photo ID capture with facial matching before entry, continuous face recognition during the exam, concurrent-login prevention, and in center-based exams biometric verification at the gate.
What is the role of question randomization in preventing malpractice?
Multiple sets and shuffled orders mean colluding candidates hold different papers, so shared answers stop mapping to shared questions — collusion loses its payoff.