Alternatives to online proctoring work in layers: assessment design that makes copying pointless, access and network scoping, browser-level controls, and post-hoc data forensics. StudyDrome Exam Manager is built on that model deliberately — it detects four classes of in-browser behaviour, adjudicates them server-side, and records no video, audio or biometric data at all.
That is the short answer. The longer one matters, because "alternatives to proctoring" is usually presented as a single swap — cameras off, something else on — and it is not. Each layer catches a different thing, and the layer with the largest effect contains no software at all.
Why are universities looking for alternatives to proctoring?
Three pressures arrived at once, and each is sufficient on its own.
Legal exposure. In August 2022 a federal district court held that Cleveland State University's practice of scanning a student's room before a remote exam was unreasonable under the Fourth Amendment, the student's privacy interest in his home outweighing the university's interest in the scan (Ogletree v. Cleveland State University, N.D. Ohio; NPR). The judgment was later vacated on procedural grounds, so it is not precedent — but the exposure it described did not leave with it.
Data-protection exposure. In 2021 Italy's data protection authority fined Bocconi University €200,000 over remote proctoring software, citing biometric processing without a legal basis and a deficient impact assessment (EDRi; Portolano Cavallo). The Court of Milan later cut the fine to €10,000, disputing the biometric classification; in May 2024 the Court of Cassation (decision 12967/2024) returned the case holding that exam-supervision technology requires a rigorous impact assessment and documented security measures.
Measurement exposure. Woldeab and Brothen studied 631 students and found that high trait test anxiety produces lower exam scores, especially for anxious students sitting exams in an online proctored setting (International Journal of E-Learning & Distance Education 34(1), 2019). A control that shifts scores differentially by trait anxiety has introduced construct-irrelevant variance: the exam is no longer measuring only what it set out to measure.
The sector has drawn its own conclusions. The University of Louisiana at Lafayette publishes an online proctoring alternatives page for its faculty, and Ohio State's Arts and Sciences distance education office publishes alternatives to proctoring exams in online classes.
What does online proctoring actually detect?
Proctoring software does not detect cheating. It detects events — a face leaving frame, a second voice, a new process — and converts them into flags for a human to review. In every system, recorded or not, adjudication is a person reading a flag.
Recording is therefore a collection strategy, not a decision strategy. It buys a rich event stream and pays in three currencies: the legal and privacy exposure above, reviewer time to watch what was collected, and a retention obligation once the recordings exist. Which tells you what an alternative has to replace — not the camera, but the evidence: enough signal to review an incident and defend a result.
What are the alternatives to online proctoring?
Five layers, ordered by how much of the problem each removes.
Layer | What it removes | What it cannot do |
|---|---|---|
Assessment design | The value of copying — an item requiring application or judgement cannot be answered from a shared file | Nothing about impersonation; a redesigned exam can still be sat by the wrong person |
Item security | The value of leakage — a large bank with randomised selection makes a leaked paper worth little | Nothing about live collusion during the sitting |
Access and network scoping | Unauthorised starts — per-candidate access codes, an IP allow-list, a supervised room | Identity assurance; a code proves possession of a code |
Browser-level controls | Casual opportunism — copy and paste blocking, fullscreen enforcement, window-leave detection | Anything on a second device, a phone, or another person in the room |
Post-hoc forensics | Undetected compromise — item statistics exposing a leaked question, conduct timelines, submission clustering | Prevention; every finding arrives after the exam |
No single row replaces proctoring. The stack does.
How do you design an exam that is hard to cheat?
The teaching-centre guidance converges on the same moves, and they are the highest-yield changes available because they alter what the exam asks rather than what it watches.
Replace recall with application. An open-book question requiring a judgement, a calculation from supplied data or a defence of a choice is not answerable from a shared answer key, and it usually maps better onto your learning outcomes than the recall item it replaced. Split one high-stakes exam into several lower-stakes components, so no single sitting justifies the risk.
Then make the paper a moving target. Randomised selection from a bank large enough to matter means two candidates rarely see the same form — and this is where item quality starts doing security work. A compromised item behaves strangely in the statistics: facility rises sharply against its history while discrimination collapses, because strong and weak candidates now answer it identically. That pattern shows up in a routine item analysis — run one on a results export today with our free item analysis tool, no account required. Item quality and item security are the same discipline, which is why writing flaw-free items belongs inside an integrity plan.
What can browser-level controls actually catch?
The honest inventory is narrower than the marketing category suggests. A browser-native layer sees four classes of behaviour: leaving the exam window, exiting fullscreen, attempting to copy question text, and attempting to paste into an answer field. Two related behaviours are caught but filed under those headings — a tab switch arrives as a window-leave event, and a blocked right-click is recorded as a copy attempt.
The hard part is not detection but not raising a false alarm, because a spurious violation on a nervous candidate's record is worse than no record at all. Three suppressions do that work in StudyDrome Exam Manager: a blur while a file-upload dialog is open does not count; a blur where the document still reports focus does not count; and on iOS a blur while a text input is focused does not count, because that is the virtual keyboard, not a departing candidate. Surviving events are debounced over 500 milliseconds before anything is recorded.
Two design decisions matter more than the detection list. Every control is off by default, chosen per exam rather than imposed on the calendar. And the browser reports while the server decides — warning thresholds and the auto-submit decision are evaluated server-side against the published exam's configuration, so a candidate cannot suppress an outcome by tampering with the client.
The limit, plainly: none of this sees the room. A phone beside the laptop, a second machine, or a person sitting off-camera is invisible to a browser. Any vendor claiming otherwise is describing a camera.
What evidence do you have if you never record anything?
A registrar or appeals panel will ask this, and the answer is concrete: an event record, which is smaller than a video and easier to defend.
Every attempt is classified by how it ended, into one of eight categories with plain labels — in progress, finished by student, time expired, test window expired, terminated for violations, finished by administrator, abandoned or session timeout, and unknown. Each attempt opens onto a timeline of its violations with timestamps and the sessions behind them, every session carrying its IP address and user agent, and the whole thing renders as a PDF for an appeal bundle. Across a sitting, each attempt is stamped with the seconds elapsed since the previous one finished, so a cluster of near-simultaneous submissions is visible to anyone reading the report.
Note what that last figure deliberately does not do: no threshold is applied and no automatic flag raised. The interval is displayed, not interpreted — the same boundary as a proctoring flag, reached with an event log instead of a recording of somebody's bedroom. Temporal clustering is a prompt to look, not a finding.
The trade is straightforward. You lose sight of the room. You gain a record a panel can read in a minute, that a candidate can be shown without exposing their home, and that carries no retention liability for video or biometric data, because none was collected. Our security overview covers how that data is handled.
When is online proctoring still the right answer?
Three cases, stated directly, because a layered approach that overreaches is just a different dishonesty.
Identity assurance for a portable credential. If the result becomes a licence or certification that travels beyond your institution, someone must establish that the person sitting the exam is the person named on it. Access codes, device binding and network scoping do not do this — they establish possession of a code and continuity of a browser session. A test centre or an identity-verifying service does.
Genuinely unsupervised, single-sitting, high-incentive exams. Where the candidate is remote and unobserved and the exam cannot be split into lower-stakes components, the design layer is unavailable and the rest carry more weight than they should.
A mandate. Some regulators and accreditors require recorded supervision. That is a compliance requirement, not a pedagogical one, and argument does not move it.
For the record: of six exam platforms surveyed on 7 August 2026, five sold webcam, AI or identity-verified proctoring. If your case is one of the three above, that market exists, use it.
A checklist for this term
- Write down what your exam actually protects against — impersonation, collusion, item leakage or opportunistic lookup. Most integrity plans conflate these, and they have different answers.
- Audit your highest-stakes paper for items answerable from a shared file, and rewrite those first.
- Check whether your bank is large enough for randomised selection to be meaningful; if not, that is this term's authoring priority.
- Run an item analysis on last cycle's results and look for the leaked-item signature — facility rising sharply against history while discrimination collapses.
- Issue access codes per candidate rather than shared, and scope by IP where the exam is sat on campus.
- Enable only the browser controls you can justify to a student, and write the justification down.
- Decide before the exam who reviews flagged attempts and against what standard. An appeal tests the adjudication policy, not the detection setting.
- Complete a data protection impact assessment. The Bocconi cases turn on that document.
Frequently asked questions
What are the alternatives to online proctoring?
Five layers, used together: assessment design that makes copying valueless, item security through large banks and randomised forms, access and network scoping such as per-candidate codes and IP allow-lists, browser-level controls like copy-paste blocking and window-leave detection, and post-hoc forensics including item statistics and submission-timing analysis. No single layer replaces proctoring; the combination does, for most exam types.
Is online proctoring legal under GDPR?
It can be, but not automatically. Italy's data protection authority fined Bocconi University €200,000 in 2021 over remote proctoring, citing processing without a valid legal basis and an inadequate data protection impact assessment. Later rulings reduced the fine and disputed the biometric classification, while affirming that exam-supervision technology requires a rigorous impact assessment and documented security measures before deployment.
Can you stop cheating in an online exam without a webcam?
You can substantially reduce it, and you cannot eliminate it — with or without a camera. Assessment design removes the payoff from copying, randomisation from a large bank removes the value of a leaked paper, and access scoping controls who starts. What no browser-based system sees is a second device or another person in the room, so identity-critical exams still belong in a supervised setting.
What does a browser-based integrity system actually detect?
Four classes of behaviour: leaving the exam window, exiting fullscreen, attempting to copy question text, and attempting to paste into an answer field. Tab switches are recorded as window-leave events and blocked right-clicks as copy attempts. It cannot see a phone, a second computer, or a person in the room, and any claim otherwise describes a camera rather than a browser.
Does removing proctoring make results indefensible at appeal?
No, provided you keep an evidence chain. A defensible bundle needs the immutable version of the paper the candidate sat, item-level statistics for the questions in dispute, a reliability figure for the exam, the standard-setting record behind the pass mark, and a per-attempt conduct record showing how the attempt ended. Video is one possible input to that chain, not the chain itself.
When should a university still use proctoring or a test centre?
When the result is a portable credential requiring identity assurance, when a high-incentive exam genuinely cannot be redesigned into lower-stakes components, or when a regulator or accreditor mandates recorded supervision. In those cases use a test centre or an identity-verifying service, and scope it to the exams that need it rather than across the whole assessment calendar.
Where to go next
If you are rebuilding an integrity policy rather than buying a product, start with the design layer: it is free, it survives every technology change, and it is the only layer that improves the exam as well as protecting it.
If you are comparing platforms, the useful question is not which one watches candidates most closely, but what evidence each can put before an appeals panel — and what it had to collect to get there. Our comparison with ExamSoft works through that difference.