Behind the Screen: What Schools Should Know About Online Proctoring Data Security

Online proctoring is usually sold as an academic integrity tool. For IT teams and privacy officers, it is also something else: a data system pointed at students during one of the most sensitive moments in school. A proctored online exam can involve test content, student identity, device activity, webcam images, screen information, timestamps, audio signals, browser behavior, flags, reports, and sometimes ID photos or room video. Once the data exists, schools have to govern it. The first security question is not 'Where is the data stored?' It is: 'Why are we collecting it at all?'

Online Proctoring Is a Chain of Data Events

Online proctoring is not one feature. It is a chain of technical events, each step creating a possible data point. Schools should stop thinking of proctoring as only a classroom-management tool. It is also an information-security and privacy workflow. If the school would ask hard questions before adopting a student information system, it should ask equally hard questions before adopting proctoring software that can collect images, recordings, identity data, and behavioral signals.

What Proctoring Tools May Monitor

• Browser monitoring: tab exits, application switches, copy-paste, fullscreen exits, blocked sites
• Screen monitoring: screen sharing, screen snapshots, or recordings during the exam
• Student webcam: face presence, multiple faces, gaze direction, environment changes
• Audio: voices or sound above a threshold level
• Identity verification: ID snapshot, face image, login token, or school account check
• Reports: raw signals converted to flags, risk scores, timelines, or teacher-facing notes

Video and Images Deserve Special Treatment

Exam video is not ordinary application telemetry. In the U.S., FERPA can apply when a photo or video is directly related to a student and maintained by the school or by a party acting for the school. Schools should know: whether video, audio, screenshots, or snapshots are captured; whether media is stored or only processed temporarily; who can view the media; whether vendors can access it; how long it is retained; and whether students or parents can request access. The safest approach is not to collect sensitive media unless the assessment actually requires it.

Data Minimization Is a Security Control

Data minimization sounds like a privacy principle. It is also a security strategy. A school cannot leak data it never collected. A vendor cannot misuse recordings that were never created. A breach cannot expose years of exam media that was deleted after a short retention period. For online proctoring, minimization should guide the default settings. A low-stakes quiz may need secure access and browser-focus signals, not camera monitoring. A final exam may justify screen sharing or optional camera checks. Collect the least sensitive data that can reasonably protect the exam.

Lockdown Browsers Solve One Problem and Introduce Others

Lockdown browsers can reduce some low-effort cheating. But they usually restrict only the device being used for the test. They do not stop a student from using a second device, printed notes, or outside help. They also raise practical security questions: students may need to install software, school devices may need centralized deployment, and IT teams need to know whether the software runs after the exam, whether it stores local data, and how students can safely uninstall it. A lockdown browser may be reasonable for some schools, but it still belongs in the normal security review process.

AI Flagging Is Not the Same as AI Judgment

Many modern proctoring systems use automated signals or AI-assisted review. From a data-governance perspective, schools should ask how flags are generated, whether confidence scores are used, whether raw evidence is stored, whether AI summaries are editable, and whether human review is required before any academic decision. Students should be told how flags work. A hidden scoring system that silently labels students as risky can damage trust quickly. The best use of AI in proctoring is to summarize noise, organize review, and help the teacher find what needs attention, not to accuse.

Vendor Privacy Policies Are Not Enough

A vendor saying 'we store data securely' is not a security review. Schools should ask for evidence. Key vendor questions: What data is collected for each proctoring mode? Is student data used for model training, product improvement, or advertising? What are the retention defaults and can schools set shorter periods? How are deletions verified? Are audit logs available? Does the vendor have SOC 2 reports, ISO certifications, or third-party security testing? What happens during a breach? The point is not to punish vendors with paperwork. It is to avoid buying a black box.

Breaches and Vulnerabilities Are Not Hypothetical

Education technology systems hold valuable data. Online proctoring systems can hold especially sensitive data because they may combine identity, contact details, exam activity, and media. A breach can expose students' personal information, weaken trust in the school, create compliance issues, and force staff into crisis response. The lesson is not 'never use proctoring software.' It is that proctoring tools need the same seriousness schools apply to other systems that handle student data. If a vendor cannot explain how it protects student data, the school should slow down.

The Best Security Feature Is Restraint

Online proctoring can help protect exam integrity. It can also collect some of the most sensitive data a school will handle. The safest strategy is not to collect everything and promise to protect it. It is to collect only what the exam requires, protect it properly, retain it briefly, limit who can access it, and make sure a human reviews the context before any decision is made. A school's goal should not be maximum monitoring. It should be trustworthy testing with the smallest necessary data footprint.