Manav.id
Comparison · 4 min read

Proctoring detects; it does not prove: assessment integrity without surveillance

Proctoring detects; it does not prove: assessment integrity without surveillance

The proctoring debate is framed as privacy versus integrity, as though institutions must choose. For a large share of assessments that framing is simply wrong, because the thing proctoring is being used to establish — that this candidate produced this submission — has a cryptographic answer that collects nothing.

Can assessment integrity be established without surveillance?

Yes, by changing what is proved. Proctoring detects suspicious behaviour and produces a probability; binding proves that the enrolled person submitted this specific work, at this time, from a credential under their sole control. One watches, the other attests.

Key takeaways
  • Proctoring bundles two questions: who is this, and are they working alone. Only the second requires observation.
  • Bound submission answers the first with a credential and collects no video, no room scan and no behavioural telemetry.
  • It does not answer the second, and any honest comparison has to say so prominently rather than in a footnote.

Separate the two questions

ProctoringCamera, screen, behaviourA suspicion scoreFalse positives on the unusualPrivacy and accessibility costBound submissionA signature over the submitted workValid or notNo false positivesCollects nothingvs
The second collects nothing and answers a narrower question definitively.

Remote proctoring is sold as one product answering one need. It is actually answering two questions with very different properties.

QuestionWhat proctoring doesWhat it achieves
Is this the enrolled candidate?ID photo comparison at session start, face monitoringProbabilistic; defeated by a determined substitution
Are they working alone and unaided?Room scan, gaze tracking, audio, screen monitoringProbabilistic; the genuine capability, and genuinely invasive

Institutions deploy the whole apparatus because it arrives as a bundle. But a substantial share of assessments — take-home work, project submissions, written assignments, portfolio work — only ever needed the first question answered, and are subjected to the second because it came in the box.

What bound submission does

The candidate enrols a credential once. At submission, they sign a canonical statement carrying a hash of the submitted work, the assessment identifier and the time.

That establishes, cryptographically, that the enrolled candidate submitted this exact artefact. It collects no video, no biometric template, no room imagery, no keystroke telemetry and no network observation. There is nothing to retain, nothing to breach and nothing to litigate about.

The honest capability comparison

This belongs in the body of the article, not in a caveat at the end.

What each approach can and cannot establish. Bound submission is weaker on the second row and that is not a small matter.
CapabilityProctoringBound submission
Candidate identity at submissionProbabilisticCryptographic
Candidate worked aloneProbabilistic observationNot addressed
No unauthorised materials usedProbabilistic observationNot addressed
Submitted artefact is unaltered after submissionNot addressedCryptographic
Data collected about the candidateExtensiveNone beyond the credential
Litigation and regulatory exposureSignificantMinimal

A candidate can enrol a credential, sign a submission, and have had someone else do the work sitting beside them. Anyone claiming otherwise is selling something.

So where does each belong?

The decision rule follows from the table rather than from a preference.

  1. Attribution only required — essays, projects, portfolios, reflective work, take-home assessments. Bound submission. Proctoring adds surveillance and no capability the assessment needs.
  2. Collusion genuinely matters — licensure examinations, gateway competency tests, high-stakes summative assessment. Proctoring, or in-person invigilation, because the second question is the point.
  3. Both matter and stakes are extreme — professional licensure. Both, with the identity question answered cryptographically so the observation can focus on what it is actually good at.

Most institutions currently apply category two tooling to category one assessments, at scale, and then are surprised by the student response.

The cost argument nobody makes

Proctoring is licensed per examination or per candidate-hour. For an institution running tens of thousands of remote assessments a term, that is a material recurring line item.

Moving category one assessments off proctoring is a cost reduction and a privacy improvement simultaneously, which is an unusual combination and is the version of this argument that survives a budget meeting.

A note on what the debate is really about

Student objection to proctoring is frequently characterised as objection to being checked. Read the litigation and the student government resolutions and it is more specific than that: objection to room scans of bedrooms, to biometric collection, to retention by a third-party vendor, and to algorithmic flagging that disproportionately misfires on some candidates.

None of those objections attaches to a credential-based attribution. That is worth understanding before assuming the resistance is to integrity measures as such.

What each approach can and cannot establish

Honest comparison
QuestionProctoringBinding
Was someone else in the room?SometimesNo
Did they use unauthorised material?SometimesNo
Did the enrolled person submit this?InferredYes
Is the submitted work unaltered since?NoYes
Cost to the honest candidateSurveillance, anxiety, exclusionOne gesture

Objections and honest limits

“Binding does not stop a proxy sitting the exam.” Correct, and stated plainly: it proves possession of an enrolled credential plus a local verification gesture, not who is in the chair. Combined with prior enrolment it makes a proxy defeat two separate processes at two separate times.

“So we still need proctoring.” For high-stakes certification, probably, and with clearer expectations of what it delivers. The argument is against relying on it as the integrity control for routine assessment, where its cost to honest candidates is highest.

Adding binding to assessment

  1. Enrol the credential separately from the exam. Ahead of time, so a proxy must defeat two processes.
  2. Sign the submission, not the session. A digest of the submitted work.
  3. Re-verify on long assessments. A mid-point gesture, seconds long.
  4. Retain the receipt with the grade record. So a later integrity question is a retrieval.

Terms used here

Remote proctoring
Monitoring a candidate through camera, screen capture and behavioural analysis during an assessment.
Submission binding
A signature over the submitted work, establishing who submitted it and that it has not changed since.
Presentation attack
Defeating an identity check by presenting synthetic or borrowed evidence to the sensor.

Frequently asked questions

Does bound submission stop cheating? No, and the article says so in the comparison table rather than burying it. It establishes who submitted the work. It does not establish who did the work.

Can a candidate lend their credential to someone else? They can hand over a device and authenticate on request, which is the same failure mode as lending an ID. It raises the cost and does not eliminate the possibility.

What about identity substitution in licensure exams? That is category two, where observation is the point. Bound submission complements proctoring there rather than replacing it.

Does this reduce our proctoring spend? Only if you move category one assessments off proctoring, which requires deciding which assessments actually need collusion detection. That classification is the work.

Does binding replace proctoring? Not for high-stakes certification. It replaces reliance on proctoring as the integrity control for routine assessment, where its cost to honest candidates is highest.

What does binding prove? That the enrolled credential submitted this specific work, with a local verification gesture, and that the work has not changed since.

Why enrol the credential ahead of the exam? So a proxy must defeat enrolment and the assessment separately, at different times, through different channels.

Where this fits in Manav

Manav binds the person to the exact record, credential or submission being authorised, and produces a receipt a receiving institution or an integrity panel can verify without contacting the issuer.

See issuance receipts →

Sources and further reading