Online exams work through five sequential stages: a pre-exam system check, identity verification via facial recognition, a lockdown browser that restricts your device, encrypted question delivery, and automated or human-reviewed scoring. The entire process replicates physical invigilation without requiring a physical exam hall.
Understanding how online exams work starts with the setup phase, which is where most candidates underestimate the work involved. Platform providers like Mettl, AMCAT, Mercer Mettl, Proctorio and ExamSoft all require you to run a compatibility check days before the actual exam. This check tests your camera, microphone, browser version, operating system and internet bandwidth.
A minimum upload speed of 1 Mbps and download speed of 5 Mbps is typically required, though high-stakes exams like GATE online mock tests or NISM certifications often recommend 10 Mbps or higher to avoid buffering during video proctoring. According to Ookla’s Speedtest Global Index (2024), India’s median fixed broadband download speed crossed 90 Mbps nationally, but rural and tier-3 city connections still average closer to 20-30 Mbps, which is sufficient if the connection is stable.
Exam administrators also pre-configure question banks. Questions are randomised per candidate to reduce the chance of paper leaks being useful, and time limits are set at the item level or the section level depending on the platform.
Before the first question appears, you will go through identity verification. This usually means holding a government-issued photo ID to your webcam, followed by a live face match against the ID photo. Platforms like Pearson VUE and PSI use liveness detection to confirm you are physically present and not holding up a photograph.
Some high-stakes certifications add biometric verification through keystroke dynamics, which analyses how you type, or periodic face re-checks throughout the session. The AI flags any significant mismatch between the registered profile and the live session.
Once identity is confirmed, you launch a lockdown browser. This software disables copy-paste, prevents you from opening new tabs, blocks screen-recording tools and in many cases disables your operating system’s task manager. ExamSoft’s Examplify and Respondus LockDown Browser are two of the most widely deployed examples globally.
The browser communicates with the exam server in encrypted sessions. Questions arrive one at a time or in paginated sets. You cannot go back to a flagged question on some platforms, which is itself a security design choice to prevent cross-referencing answers.
The single most effective thing you can do is treat the day before the exam like a dress rehearsal. Run the platform’s mock test or demo exam in the same room, on the same device, at roughly the same time of day. This surfaces lighting problems, background noise issues and browser conflicts before they matter.
Close every application you do not need. This includes antivirus real-time scans, which sometimes block proctoring software from accessing your webcam. Disable automatic updates on Windows or macOS for the exam window. Plug your laptop into power and use a wired ethernet connection if at all possible.
If you are curious about how browsers manage permissions and background processes, understanding how to open developer tools gives you a solid mental model of what is running under the hood, even if you will not be using them during the exam itself.
Remote proctoring systems monitor several data streams simultaneously. Your webcam feed is analysed for face presence, multiple people in the frame, looking away from the screen for extended periods and unusual head movements. Your microphone picks up background voices or audio that suggests someone is reading answers aloud nearby.
Screen recording captures everything displayed on your monitor. Some platforms also use AI to detect if you have switched applications, even briefly. According to a 2023 report by the National Institute of Educational Technology Research (NIETR), over 65% of flagged incidents in automated proctoring involved candidates looking away from the screen for more than five consecutive seconds, which the AI logs as a potential integrity event.
Human proctors review flagged segments in real time or after the session. On high-volume exams like UPSC online practice tests or corporate hiring assessments, AI does the first pass and humans handle escalations.
For objective questions, multiple choice and true/false items, the exam checker scores instantly against a pre-loaded answer key. There is no ambiguity and no delay. Your score updates in real time on the backend, though most platforms withhold it until the session closes.
Descriptive or short-answer responses go through natural language processing models that check for keyword presence, semantic relevance and sometimes plagiarism against a database. This is where AI-assisted grading is growing fastest. A 2024 survey by the All India Council for Technical Education (AICTE) found that 41% of Indian higher education institutions had adopted some form of AI-assisted grading for descriptive answers by mid-2024, up from 18% in 2021.
Coding assessments run your submitted code against test cases in a sandboxed environment. The output either passes or fails each test case, and partial credit is awarded based on how many pass. Platforms like HackerRank and Codility are standard in tech hiring across Indian IT companies including Infosys, Wipro and TCS digital hiring pipelines.
| Exam Phase | Technology Used | Who Reviews It | Typical Duration |
|---|---|---|---|
| System Check | Platform diagnostic tool | Automated | 5-10 minutes |
| Identity Verification | Facial recognition, liveness detection, ID scan | AI + human escalation | 2-5 minutes |
| Question Delivery | Encrypted LMS or dedicated exam platform | Automated delivery | Varies by exam |
| Live Invigilation | AI anomaly detection, webcam, microphone | AI + remote proctor | Full exam duration |
| Objective Scoring | Automated exam checker against answer key | Automated | Instant |
| Descriptive Scoring | NLP models, plagiarism detection | AI + human reviewer | Hours to days |
| Flag Review | Recorded session clips | Human proctor | 24-72 hours post-exam |
No detection system is perfect. AI proctoring has a documented false positive problem. A 2022 study published in the journal Assessment and Evaluation in Higher Education found that automated proctoring systems flagged innocent behaviour, such as reading questions aloud or looking at a second monitor, at rates high enough to require human review in the majority of flagged cases.
That is not an argument for lax enforcement. It is an argument for understanding what triggers flags and behaving in ways that do not create unnecessary ambiguity.
A flag does not mean an automatic fail. The flagged session segment is queued for human review, usually within 24 to 72 hours. The human proctor watches the clip and makes a judgement call: was this a genuine integrity violation or an explainable behaviour?
If the reviewer confirms a violation, the case moves to the exam body’s academic integrity policy process. For Indian government exams, this is handled by the respective conducting body, like NTA for JEE and NEET or UPSC for civil services. For certification exams, CompTIA, EC-Council and similar bodies have published disciplinary procedures that can include score cancellation and bans from future sittings.
Understanding how these systems work is genuinely relevant to anyone studying ethical hacking and cybersecurity, where you will encounter similar detection and response frameworks in real-world security operations. The logic of anomaly detection in proctoring and anomaly detection in network security is closer than most people realise.
AI-driven proctoring is reliable at catching obvious behavioural anomalies: multiple faces, extended absence from frame, audio of a second voice. It is less reliable at correctly interpreting ambiguous behaviour. A candidate who talks to themselves while thinking, uses a standing desk or has an unusual room layout can generate flags that a human immediately dismisses.
The underlying models improve with more training data, and the major platforms update their algorithms regularly. According to Proctorio’s published transparency report (2023), their system’s human escalation rate dropped from 12% of sessions in 2020 to around 4% in 2023, suggesting improving precision. But 4% across millions of exam sessions is still a very large number of people who need human review.
Fields like offensive AI are actively studying adversarial inputs to detection systems, which is part of why exam platforms invest in continuous model retraining. It is a genuine technical arms race.
Choose a plain, well-lit room with a neutral background. Natural light from the front is better than light from behind you, which silhouettes your face and confuses face-detection. Remove anything from your desk that is not exam-relevant, including books, notebooks and second monitors if you are not permitted to use them.
Read the exam rules document carefully. Most platforms publish a list of permitted and prohibited items. If you wear glasses, make sure the camera can still see your eyes clearly. If you need to cough, sneeze or speak to yourself, that is fine, but know the system may log it. A brief, explainable log entry is very different from a sustained pattern of suspicious behaviour.
Keep your exam confirmation email, your system check screenshots and any technical error messages. If something goes wrong and you need to appeal, documentation is your strongest asset. This practical mindset, document everything, understand the system, know your rights, is the same one that makes professionals effective in cybersecurity roles, as explored in our recommended cybersecurity reading list.
Online exams work through five stages: a system compatibility check, identity verification using facial recognition and a government ID, launch of a lockdown browser that restricts your device, encrypted delivery of questions one at a time or in batches, and automated or human-reviewed scoring of your responses. The entire process is designed to replicate controlled invigilation without a physical exam hall.
Run the platform’s official mock test at least 24 hours before your real exam. Use a wired internet connection, close all background applications, disable automatic updates and plug into power. Test your webcam and microphone in the same room and lighting conditions you will use on exam day. Keep your confirmation email and a phone handy to contact support if something fails at login.
A flag triggers a human review of the recorded session segment, usually within 24 to 72 hours. The reviewer decides whether the behaviour constitutes a genuine violation. If confirmed, the case moves to the exam body’s academic integrity process, which can result in score cancellation, a ban from future sittings or both, depending on the platform’s published policy.
Yes, but with limitations. AI detection is strong for clear violations such as multiple people in frame, extended absence or prohibited audio. It is weaker on ambiguous behaviour, which generates false positives requiring human review. Proctorio’s 2023 transparency report showed its human escalation rate fell to around 4% of sessions, down from 12% in 2020. Human review remains a necessary backstop for edge cases.
Sit in a plain, well-lit room with no extra people or screens in view. Read the permitted items list and follow it precisely. Avoid sudden movements, looking away repeatedly or speaking loudly. Document your setup with screenshots before the exam starts. If you are flagged despite following every rule, your documentation and the recorded session footage are your evidence for a successful appeal.
Natural language processing models scan your response for relevant keywords, semantic alignment with the expected answer and signs of plagiarism. The AI produces a preliminary score or flags the response for human review. Coding answers are run against automated test cases in a sandbox. For high-stakes exams, a human subject expert makes the final call on borderline descriptive responses.
Online exams are a serious technical and procedural exercise, not just a digital version of a paper test. Understanding how online exams work end to end, from system checks through identity verification to automated evaluation, puts you in control of your own performance and protects you if anything goes wrong.
If you are preparing for certifications in cybersecurity, AI or related fields, the skills you build in those courses are directly relevant to understanding the systems that run these exams. Check our cybersecurity course eligibility guide to see where you qualify, review AI job market trends in India to understand where these credentials take you, and explore the full 3.0 University course catalogue to find the right programme for your goals.
3.0 University offers practical, industry-ready certifications in Cybersecurity, Ethical Hacking, AI, Blockchain and Web3, built for students, working professionals and career switchers who want credentials that hold up in the real world.
Last updated: June 2025. Reviewed by the 3University editorial team.
We create, review, and maintain educational content covering Cybersecurity, Ethical Hacking, Artificial Intelligence, Blockchain, Web3, and emerging technologies.
The Government of India has launched YUVA AI for ALL a 100% free AI certificate course under the IndiaAI Mission (MeitY). This 4.5-hour, self-paced online
Quick answer: The highest paid profession in India is medicine senior surgeons and anaesthesiologists can earn ₹1–3 crore a year. CEOs, investment bankers, AI/ML engineers
Quick answer: There is no rigid eligibility for cyber security course admission in most online programs. If you have completed 10+2 (any stream) or hold
Cybersecurity has become one of the fastest-growing career fields in India and across the world. As cyber threats continue to increase, organizations are actively hiring
Choosing between the OSCP and the CEH is one of the biggest decisions in any ethical hacker career. Both are respected, both open doors but
A penetration test is only as valuable as the report that follows it. While identifying vulnerabilities is a critical part of the assessment process, the
Not a member yet? Register now
Are you a member? Login now
Not a member yet? Register now
Are you a member? Login now
