ExamPlatform’s AI proctoring system runs continuously from the moment a student starts their exam until the session ends — without interrupting the student’s flow. Rather than pausing the exam or displaying intrusive alerts, the system silently captures integrity signals in the background, aggregates them into a risk score, and surfaces a full timeline of flagged events for examiners and administrators to review after the exam. The philosophy is to keep the student focused on answering questions while ensuring that the integrity record is complete and available for human review.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/Saurabh-07586/examplatform/llms.txt
Use this file to discover all available pages before exploring further.
Monitored Signals
The proctoring engine tracks the following signals throughout the session:- Gaze away events — The eye-tracking model detects when the student’s gaze direction moves away from the screen. Each episode is timestamped and added to the suspicious activity timeline.
- Face not detected — When the webcam feed no longer contains a recognisable face (the student has moved out of frame, covered the camera, or the lighting has failed), the system logs a face-missing event.
- Multiple faces — If more than one face is detected in the webcam frame at the same time, the event is flagged and counted. This may indicate that a second person is assisting the student.
- Mobile detection — The model scans the webcam frame for the presence of a secondary device (a smartphone or tablet). Each detection is logged with a timestamp.
- Tab switches — If the student navigates away from the exam window — switching to another browser tab, another application, or the operating system desktop — the event is captured and a running count is maintained. The count is visible to the student on their monitoring panel and to examiners in the proctoring report.
- Fullscreen exits — The exam runs in mandatory fullscreen mode. Every time the student exits fullscreen, the exit is logged and the count increments. Students are informed that this is recorded before the exam begins.
- Connectivity issues — Network interruptions are detected through the heartbeat signal that each active session sends to the server. Gaps in the heartbeat are flagged as connectivity issues and included in the integrity timeline.
AI Risk Score
All of the signals above are combined into a single composite AI Risk Score that gives a high-level summary of the session’s integrity posture. The score is classified into three levels:| Level | Meaning |
|---|---|
| Normal | No significant integrity signals detected. The session appears clean. |
| Elevated | One or more moderate signals have been detected (e.g. a small number of gaze-away events or a tab switch). The examiner should review the timeline. |
| Critical | Multiple or severe integrity signals have been detected (e.g. multiple faces, repeated mobile detections, frequent fullscreen exits). Immediate examiner review is recommended. |
Student Monitoring Panel
During the exam, a collapsible AI Monitoring panel is visible on the student’s interface. It provides the student with real-time feedback on their session’s integrity status so they can correct inadvertent issues (such as moving back into frame if face detection is lost). The panel displays:- Webcam — Live camera feed with a “Recording” indicator
- Microphone — Active / Unavailable status
- Eye tracking — Active / Focused status
- Face detection — Clear / Flagged status
- Mobile detection — Detected count
- Tab switches — Running count of tab-switch events
- AI risk score — Normal / Elevated / Critical
- Internet — Stable / Unstable / Poor
- Battery — Current battery level of the device
Examiner and Admin: AI Proctoring Reports
After a session ends, examiners and administrators can access the AI Proctoring Reports section. Each report includes:- Overall exam integrity status — A summary banner showing the session’s final AI risk score
- AI Risk Score — The composite score (Normal / Elevated / Critical) for the session
- Suspicious activity timeline — A chronological log of every flagged event with timestamps, making it possible to correlate events with specific points in the exam
- Per-event breakdown — Individual counts for each signal category:
- Gaze away events
- Face not detected
- Multiple faces
- Mobile detections
- Tab switches
- Fullscreen exits
AI proctoring findings are advisory. The system surfaces evidence to support the examiner’s judgment — it does not automatically disqualify or penalise students. The examiner always has the final say when reviewing integrity reports and deciding whether a flagged event represents a genuine integrity concern or a benign technical issue.
Force Submit (Admins — Live Monitoring)
Administrators have the ability to Force Submit any active student session directly from the Live Monitoring dashboard. This is available for situations where an admin determines that a session must be terminated immediately — for example, in response to a Critical risk score combined with confirmed suspicious behaviour. Force Submit immediately closes the session and records all answers saved up to that point, identical to a normal auto-submission.Browser Requirements
ExamPlatform’s proctoring system relies on specific browser capabilities. Students should be aware of the following requirements and restrictions before starting an exam:- Fullscreen mode is mandatory. The exam interface enters fullscreen automatically when the session starts. Exiting fullscreen is logged as an integrity event and the student is prompted to re-enter fullscreen.
- Copying, pasting, right-clicking, and printing are disabled for the duration of the exam. These actions are intercepted at the browser level to prevent content from being copied out of or into the exam window.
- A browser with camera and microphone API support is required. Students using browsers that do not support these APIs will see a warning and will be unable to start the exam.
- A stable internet connection and a charged device are strongly recommended. Network drops are logged as connectivity issues and battery failure mid-exam will disrupt the session.