{"id":178122,"date":"2025-08-29T12:24:07","date_gmt":"2025-08-29T11:24:07","guid":{"rendered":"https:\/\/www.network-it.be\/ongecategoriseerd\/how-chatgpt-abuse-in-physician-entrance-exam-could-have-been-prevented-an-it-and-cybersecurity-perspective\/"},"modified":"2025-12-28T20:47:51","modified_gmt":"2025-12-28T19:47:51","slug":"how-chatgpt-abuse-in-physician-entrance-exam-could-have-been-prevented-an-it-and-cybersecurity-perspective","status":"publish","type":"post","link":"https:\/\/www.network-it.be\/en\/information\/how-chatgpt-abuse-in-physician-entrance-exam-could-have-been-prevented-an-it-and-cybersecurity-perspective\/","title":{"rendered":"How ChatGPT abuse in physician entrance exam could have been prevented – an IT and cybersecurity perspective"},"content":{"rendered":"\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

In July 2025, it was first found that students used ChatGPT during the medical entrance exam in Flanders, despite security measures such as disabling Internet access<\/span>. This scandal brought acute attention to exam integrity and raises important questions about the robustness of technical and operational safeguards, especially in temporarily set-up exam environments where IT resources are limited. In this blog, IT and cybersecurity experts highlight the risks of AI-assisted fraud, evaluate existing deficiencies and explain how organizations can ensure integrity with smart measures, even in small-scale exams. You’ll learn what technologies and processes, at a minimum, are necessary to effectively keep ChatGPT and similar systems out of physician training access depots.<\/span><\/p>

1. Context: the incident and the cybersecurity challenge<\/h3>

In July 2025, three candidates were caught using ChatGPT during the nursing and medicine exam despite the fact that exam booths should have prevented Internet access<\/span>. This proves that even with supposedly layered security, vulnerabilities remain. In addition, the exam board is currently investigating objections from other students about possible mass fraud and a remarkably high pass rate (47%, versus 19% in 2024)<\/span>. Such incidents demonstrate that exam environments need robust cybersecurity design, even when resources are limited.<\/span><\/p>

2. Technical measures against AI-based fraud<\/h3>

2.1 Secure (lockdown) browser & screen monitoring<\/h4>

  • Use a secure browser that blocks all access to unauthorized websites, tabs and apps<\/span>.<\/p><\/li>

  • Screen captures and screenshots allow suspicious activity to be captured automatically<\/span>.<\/p><\/li>

  • Platforms such as Exam.net prevent copy\/paste and detect focus-switching<\/span><\/p><\/li><\/ul>

    2.2 AI-assisted proctoring and live monitoring<\/h4>

    • Tools such as Proctor360 offer cameras with 360\u00b0 views, multi-camera setups, AI behavioral analysis and live intervention capabilities<\/span>.<\/p><\/li>

    • Talview combines secure browser, room check from 360\u00b0, behavioral analysis and proctor logs<\/span>.
      This hybrid approach reduces risk of fraud as well as enabling human grip action.<\/span><\/p><\/li><\/ul>

      2.3 Demand design and multimodality<\/h4>

      • Avoid reusing questions. Create unique, ever-changing exam questions-so ChatGPT has less opportunity to anticipate<\/span>.<\/p><\/li>

      • Insert images and drawings into questions, including embedded text labels; such multimodal questions have been found to significantly weaken ChatGPT<\/span>.<\/p><\/li>

      • Use question sets with sequential building blocks that test comprehension and frustrate AI-generated answers.<\/span><\/p><\/li><\/ul>

        2.4 Question-fulnerability scoring<\/h4>

        • Investigate which question types ChatGPT answers least well. Use NLP analysis to measure this by question type and avoid vulnerable questions. <\/span><\/p><\/li><\/ul>

          2 3. Operational approach for temporary and small-scale sites <\/h4>

          Even in those situations, at least the following measures can be applied:<\/span><\/p>

          • Preliminary room and device check<\/strong>: well in advance, scan the room and table for hidden devices. A security officer performs this check. <\/span><\/p><\/li>

          • Mobile proctoring kit<\/strong>: a tablet or smartphone with proctoring app (with secure browser + video stream) is already sufficient for small exam cohorts.<\/span><\/p><\/li>

          • Hybrid proctoring<\/strong>: combine live camera monitoring with an attendant; for balance between cost and security.<\/span><\/p><\/li>

          • Log and video usage<\/strong>: keep logs of device usage, behavioral data, video recordings and suspicious events.<\/span><\/p><\/li>

          • Backup paper exams\/pen-and-paper<\/strong>: if technology fails, can quickly switch to traditional forms of exam.<\/span><\/p><\/li><\/ul>

            2 4. Prevention through processes and awareness <\/h4>

            • Oral or spoken testing (viva voce)<\/strong>: as applied by the University of South Australia, rapid oral questioning of less than 20 minutes can greatly inhibit fraud<\/span>.<\/p><\/li>

            • Behavioral analysis and forensic log analysis<\/strong>: an AI human-in-the-loop system helps detect patterns of exam fraud<\/span>.<\/p><\/li>

            • Cultural shift to digital ethics<\/strong>: educate candidates about AI tools, their impact and while promoting fair use; it may help as suggested by Reading study on AI-assisted deception.<\/span><\/p><\/li><\/ul>

              3. AI-cheat-proof exam framework: 6 layered layers of defense<\/h3>

              3.1 Layer 1: Identity authentication & access management<\/h4>

              The first layer of defense must absolutely guarantee that the right candidate is taking the exam – and that he\/she is not switching or receiving outside help.<\/p>

              Technical measures:<\/strong><\/p>

              • Multi-Factor Authentication (MFA)<\/strong> at login: password + mobile OTP code.<\/p><\/li>

              • Biometric verification (face match)<\/strong>: used before and during the exam to validate identity.<\/p><\/li>

              • Geofencing<\/strong> via GPS\/IP logging: prevents candidates from connecting from outside approved locations.<\/p><\/li>

              • Randomization schemes<\/strong> for seat or login times to make collusion more difficult.<\/p><\/li><\/ul>

                Relevance to temporary sites:<\/strong><\/p>

                • Easily deployable via cloud-based tools (such as ProctorExam or Talview).<\/p><\/li>

                • Requires only basic cameras + internet; no heavy infrastructure.<\/p><\/li><\/ul>

                  3.2 Layer 2: Device and system hardening<\/h4>

                  Once identity is confirmed, the device on which the exam is taking place must be under full control. AI interfaces such as ChatGPT must be inaccessible. <\/p>

                  Technical measures:<\/strong><\/p>

                  • Secure lockdown browser<\/strong>: blocks access to Internet, apps, keyboard shortcuts, copy-paste, print screen.<\/p><\/li>

                  • Completely closed OS profile (kiosk mode)<\/strong>: only the exam application is accessible.<\/p><\/li>

                  • MDM policy for temporary devices<\/strong>: checks for USB injections, Bluetooth connections, microphone usage.<\/p><\/li>

                  • BIOS\/UEFI security<\/strong>: prevents booting via USB or network drive.<\/p><\/li>

                  • Whitelisting of network traffic<\/strong>: only traffic to the exam server is allowed.<\/p><\/li><\/ul>

                    Application to temporary exam rooms:<\/strong><\/p>

                    • Use of preconfigured laptops based on “immutable images”.<\/p><\/li>

                    • Network segregation per device via VLAN or private WiFi SSID per row.<\/p><\/li><\/ul>

                      3.3 Layer 3: Environmental security & infrastructure control<\/h4>

                      In addition to the device, the candidate’s physical and digital environment must also be rigorously monitored to rule out hidden AI tools or spy devices.<\/p>

                      Technical measures:<\/strong><\/p>

                      • RFID, Bluetooth, and Wi-Fi scanning<\/strong> for unauthorized signals (smartwatches, mobile hotspots).<\/p><\/li>

                      • Use of Faraday devices<\/strong> (signal blocking): for example, domes over desks or shielded classrooms.<\/p><\/li>

                      • 360\u00b0 video surveillance<\/strong> or dual camera setup<\/strong> (frontal + overhead).<\/p><\/li>

                      • Ambient noise monitoring<\/strong> to detect voice assistants (such as Siri or Google Assistant).<\/p><\/li><\/ul>

                        For temporary spaces:<\/strong><\/p>

                        • Portable scanners (such as Flipper Zero or Wi-Spy Air) are ideal for on-site pre-scan.<\/p><\/li>

                        • Backup power & failover internet for surveillance cameras.<\/p><\/li><\/ul>

                          3.4 Layer 4: Content AI resistance of exam questions<\/h4>

                          Even if candidates had access to AI, smart question design can prevent those tools from generating useful answers.<\/p>

                          Technical & didactic strategies:<\/strong><\/p>

                          • Multimodal questions<\/strong> with pictures, diagrams or charts in which AIs perform weaker.<\/p><\/li>

                          • Cumulative question sets<\/strong>: where each answer depends on prior understanding.<\/p><\/li>

                          • Context-sensitive assignments<\/strong> (e.g., clinical cases with conflicting information).<\/p><\/li>

                          • Dynamic randomization<\/strong>: no two candidates receive the exact same questions.<\/p><\/li><\/ul>

                            Support techniques:<\/strong><\/p>

                            • AI vulnerability scoring: test which questions are easily solved by LLMs.<\/p><\/li>

                            • Use tools such as OpenAI Detector, GPTZero for pre-screening queries.<\/p><\/li><\/ul>

                              3.5 Layer 5: Monitoring & behavioral analysis during examination<\/h4>

                              Live detection of suspicious patterns is essential for real-time action.<\/p>

                              Technical resources:<\/strong><\/p>

                              • Live video & screenshare monitoring<\/strong>, combined with AI analysis of facial expressions, eye movement, and mouse\/key interactions.<\/p><\/li>

                              • Input monitoring<\/strong>: detection of inhuman typing rates or patterns (burst typing).<\/p><\/li>

                              • Browser focus analysis<\/strong>: records when the candidate clicks out of the exam window.<\/p><\/li>

                              • Audit logging<\/strong>: central recording of all system and user interactions.<\/p><\/li><\/ul>

                                For scalable temporary setups:<\/strong><\/p>

                                • Cloud-based proctoring platforms with integrated analytics.<\/p><\/li>

                                • Lightweight clients (Chromebook, tablets with external camera).<\/p><\/li><\/ul>

                                  3.6 Layer 6: Post-exam forensic evaluation & anomaly detection<\/h4>

                                  After the exam, automatic analysis should detect anomalous patterns and escalate potential frauds.<\/p>

                                  Techniques:<\/strong><\/p>

                                  • Natural Language Forensics<\/strong>: recognizes output that resembles LLM language structure (repetitions, style).<\/p><\/li>

                                  • Statistical comparison models<\/strong>: comparison of individual scores with group mean.<\/p><\/li>

                                  • Plagiarism & duplication control<\/strong> on answer structures.<\/p><\/li>

                                  • Behavioral forensics<\/strong>: comparison with historical behavioral pattern of the same candidate.<\/p><\/li><\/ul>

                                    Temporary locations:<\/strong><\/p>

                                    • Data automatically sent to central server for analysis.<\/p><\/li>

                                    • Low latency required to synchronize video and input data in a timely manner.<\/p><\/li><\/ul>

                                      Conclusion<\/h3>

                                      The observation of ChatGPT use during the medical entrance exam in Flanders shows that technical as well as organizational security layers are crucial-even in small, temporary exam environments. By deploying a block of secure browsers, proctoring (AI assisted and human), multimodal question design, behavioral analysis and clear processes, we significantly reduce the risk of AI-based fraud. Add digital ethics and oral review for maximum robustness. Want support implementing a secure exam environment? Then schedule a call to discuss how Network IT can secure your exam processes against ChatGPT abuse. <\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t

                                      <\/div><\/div><\/div><\/div>","protected":false},"excerpt":{"rendered":"

                                      In July 2025, it was first found that students used ChatGPT during the medical entrance exam in Flanders, despite security measures such as disabling Internet access. [\u2026]<\/span><\/p>\n","protected":false},"author":1,"featured_media":157259,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"content-type":"","wds_primary_category":0,"footnotes":""},"categories":[63,62,54,60],"tags":[],"class_list":["post-178122","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-cybersecurity","category-information","category-security"],"_links":{"self":[{"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/posts\/178122","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/comments?post=178122"}],"version-history":[{"count":1,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/posts\/178122\/revisions"}],"predecessor-version":[{"id":178127,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/posts\/178122\/revisions\/178127"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/media\/157259"}],"wp:attachment":[{"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/media?parent=178122"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/categories?post=178122"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.network-it.be\/en\/wp-json\/wp\/v2\/tags?post=178122"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}