Exam integrity and anti-cheating software for high-stakes assessment

The cheapest integrity control is the one applied before the exam opens. Drawing each candidate's paper from a structured item bank means neighbours are not looking at the same question in the same order, and a leaked screenshot from the first sitting does not describe the second.

Randomisation works at several levels — which items are drawn, what order they appear in, and how the options within each item are ordered — so equivalent papers stay comparable in coverage and difficulty while differing in surface form. Where item banks need to grow fast enough to support this, AI Questions generates across 15 question types and 18 languages.

Randomisation limits what a candidate gains from a neighbour; lockdown limits what they gain from the device in front of them. The exam session is delivered into a constrained environment where navigating away, opening other applications or reaching outside the assessment is restricted rather than merely discouraged.

Environmental controls matter most at scale, where a small percentage of attempts becomes a large absolute number. iTest is built for sittings of 120,000+ concurrent candidates and 2M+ exams per year, so lockdown behaviour has to hold under national examination load, not just in a computer lab of thirty.

Every other control assumes the right person is in the chair. In remote and distributed sittings that assumption is the weakest link, and it is the one an appeal will attack first — particularly for licensing and certification exams where the credential has real economic value. iTest runs Egypt's EMLE medical licensing examination, sat by 9,397 doctors.

Identity verification links the session to an enrolled candidate at the start and keeps that link accountable through the sitting, so the record does not merely show that an exam was completed but that a specific candidate completed it. Where the institution already runs its own identity provider, single sign-on avoids creating yet another credential to share or lose.

A pile of flags is not evidence. Left unweighted, it produces two failures at once: reviewers drown in false positives from candidates who simply looked away, and genuine irregularities are lost in the noise. What a review board needs is a defensible judgement with its reasoning attached.

iTest consolidates session signals into an integrity score, so sessions are triaged by how anomalous they actually are and human review is spent where it changes an outcome. iTest reports a 94% integrity score and 99%+ AI accuracy — so the grade under dispute and the integrity assessment beside it are both traceable rather than asserted.

Assume the exam will be challenged. The question is whether the institution can reconstruct the session months later: which form the candidate received, when the session opened and closed, what the environment did, which signals were raised, who reviewed them, what they decided and on what basis. A timestamped record that shows any later change is what separates a defensible decision from a contested one.

That record also protects candidates. A flagged session that review clears leaves a documented clearance rather than a rumour, and an appeals committee reviewing evidence — rather than adjudicating between two accounts — reaches a decision that survives scrutiny. For institutions that cannot place examination data with a third party, iTest runs cloud, on-premise or hybrid, and is built against GDPR and FERPA.