Compromised-item detection tries to spot items that have been leaked, overexposed, or otherwise become too predictable in computerised adaptive testing. It matters because adaptive systems can feel secure while still being vulnerable to item exposure, pre-knowledge, and repeated misuse. The strongest reading of the sources is that detection is useful, but only when paired with exposure control, item-bank governance, and careful review of how items circulate.

Compromised-item detection is the statistical or forensic identification of test items that may no longer be secure because candidates have seen them before, shared them, or learned them through unauthorised means. In CAT, this is especially important because item selection is dynamic, so a compromised item can do more damage than in a fixed form if the system does not notice quickly.

Adaptive testing is often promoted as efficient and psychometrically elegant, but efficiency can be fragile if security is not managed with equal care. If item exposure is not monitored, the system can drift into predictable patterns that candidates or coaching services can exploit.

- **Item compromise**: an item becomes known to candidates before or during legitimate use. - **Pre-knowledge**: advance access to item content. - **Item exposure**: how often an item is administered. - **Adaptive selection**: choosing the next item based on the candidate’s responses. - **Security signal**: a statistical indicator that an item may need review or retirement.

The source suggests that the main value of compromised-item detection is in protecting both security and fairness. If compromised items stay live, candidate scores can be distorted in favour of those with unauthorised access.

The open question is how much detection alone can achieve. A good detection model can identify suspicious items faster, but it does not by itself stop exposure, control item sharing networks, or settle whether a flagged item should be retired, recalibrated, or reviewed by content experts.

- overexposed items becoming easier to coach - false alarms if normal statistical variation is mistaken for compromise - under-detection if the model lags behind actual leakage patterns - relying on post-hoc detection instead of item-bank discipline

1. Track item exposure as a routine security measure. 2. Use compromise detection to prioritise human review. 3. Retire or refresh items that keep appearing in suspicious patterns. 4. Combine detection with item-bank rotation and form diversity. 5. Recheck whether security controls still fit the CAT design.

| Approach | Strength | Limitation | |---|---|---| | **Detection only** | Surfaces suspicious items quickly | Can be too late if exposure is already widespread | | **Detection plus exposure control** | Better balance of security and efficiency | Requires stronger governance and analytics | | **Broad redesign of item-bank strategy** | Reduces dependence on any single item | More expensive and operationally demanding |

A licensure programme sees odd performance on a small cluster of CAT items. Rather than assuming the entire test is compromised, it runs item-level analysis, checks exposure history, and retires the most suspicious items while refreshing the bank. That is a practical security response because the risk sits in the item lifecycle, not just in the candidate session.

- Research article note on a model for detecting compromised items in CAT. - TCN source note on exposure control and form diversity in test assembly.

This is more of a psychometric and security-methods topic than a product category. Suppliers may offer item analytics or forensics, but the central design issue is still how item banks are monitored and governed.

### Is compromised-item detection the same as cheating detection? No. It focuses on item security and exposure, not every form of misconduct. ### Why is this especially important in CAT? Because adaptive delivery can reuse item content in ways that make exposure patterns harder to see unless they are monitored deliberately. ### Can detection replace item-bank rotation? No. Detection helps, but item rotation and exposure control still do most of the preventive work.

Tim Burnett (Admin)

`Test Community Network. "Compromised item detection in computerised adaptive testing." TCN Wiki. Last reviewed 2026-05-05. https://www.testcommunity.network/wiki/test-security-compromised-item-detection-cat`

- Research article note on a model for detecting compromised items in CAT. - TCN source note on exposure control and form diversity in test assembly.