Closing the Benchmark Gap in Browser-Extension Proctoring: A Formally Specified, Reproducibly Evaluated Architecture for Real-Time, BYOD-Compatible Examination Malpractice Detection

In the modern world of education, the shift from paper exams to computer-based tests has brought undeniable efficiency, but it has also opened a door to a new kind of academic dishonesty that leaves no physical trace. In a traditional exam hall, a proctor can see if a student is looking at a hidden cheat sheet or whispering to a neighbor. On a computer, however, a student can simply click away from the exam window to check a forbidden website, copy a question, or switch to a different browser tab, all while the exam screen still appears to be open and active. This is known as "context switching," and it has created a difficult dilemma for schools. To stop it, institutions have traditionally relied on heavy-handed software that locks down the entire computer, preventing access to anything but the exam. While effective, this approach requires administrative rights that students often do not have on their own personal devices, making it impossible to use in "Bring Your Own Device" policies where students take their own laptops to class. Other solutions, like using webcams to watch students, demand high internet speeds and raise serious privacy concerns.

Researchers at the Air Force Institute of Technology in Kaduna, Nigeria, have now stepped into this gap with a new approach that does not try to lock the computer down, but rather watches the browser window itself. They built and tested a lightweight software add-on, or extension, that runs inside the web browser during an exam. Unlike the heavy software that takes over the whole machine, this tool operates quietly within the browser, monitoring for specific actions like opening a new tab, minimizing the exam window, or copying text. The team did not just build this tool; they subjected it to a rigorous, controlled test to measure exactly how well it works, how fast it reacts, and how much power it uses. Their goal was to move beyond the marketing claims of commercial companies and provide schools with hard, independent data on whether this technology is reliable enough to trust with high-stakes exams.

The researchers developed a system they call the Host-Based Malpractice Detection System, which functions as a digital watchdog inside the browser. When a student starts an exam, the system begins monitoring every movement within that specific browser window. If the student tries to switch to a different tab, the system catches the change immediately. If they try to open a developer tool to inspect the code, or if they attempt to copy and paste exam content, the system flags the action. Crucially, the system also watches for a specific type of silence: if the connection between the student's browser and the exam server is broken, the system assumes the student might have tried to disable the monitoring. To handle this, the software sends a tiny, regular signal, like a heartbeat, to the server every ten seconds. If the server stops receiving these signals for a set period, it alerts a human proctor that something is wrong. This design allows the system to work on any personal computer without needing special installation rights, making it compatible with the "Bring Your Own Device" policies that many schools now prefer.

To see if this system actually works in the real world, the team ran a series of automated tests that simulated an academic dishonesty environment. They created a scenario where fifty students were taking an exam at the same time, and they programmed scripts to perform five hundred different unauthorized actions, such as switching tabs, minimizing windows, or cutting and pasting text. The system was remarkably effective at catching these attempts. It successfully detected unauthorized tab switches and clipboard actions one hundred percent of the time. When it came to window minimization, where a student tries to hide the exam window, the system caught the vast majority of attempts, missing only three out of one hundred and fifty tries. These few misses happened only when the student switched windows so quickly that the browser's internal processing could not register the change in time. Overall, the system achieved a detection accuracy of ninety-nine point five percent for actions happening inside the browser.

Speed was another critical factor the researchers measured. In a live exam, a warning needs to reach the proctor almost instantly to be useful. The team found that when a violation occurred, the system sent an alert to the administrator's dashboard in an average of four hundred and fifty milliseconds over a standard Wi-Fi connection. This is less than half a second, meaning a proctor would see a warning almost the moment a student tried to engage in academic dishonesty. The system also proved to be very gentle on the computer's resources. While older, heavy-duty lockdown software can use up nearly half of a computer's processing power, this new browser extension used less than eight percent of the CPU and required very little memory. This low resource footprint is what makes it possible to run on older student laptops without slowing them down, a key advantage over the more invasive alternatives.

The researchers also investigated how to set the "heartbeat" timer, which determines how long the system waits before assuming a connection has been lost. If the timer is set too short, the system might falsely accuse a student of academic dishonesty just because their Wi-Fi flickered for a moment. If it is set too long, a student could disable the software and engage in academic dishonesty for a long time before anyone notices. By testing different time limits, the team discovered that a forty-five-second wait time was the sweet spot. This duration provided a safety margin that was long enough to ignore normal, brief Wi-Fi hiccups, resulting in a false alarm rate of only two point one percent, while still being short enough to catch a deliberate attempt to disconnect quickly. This finding suggests that schools should not use a single, fixed setting for this timer everywhere, but rather adjust it based on the specific quality of their internet connection.

Despite these impressive results, the researchers are careful to define exactly what their system can and cannot do. The system is designed to watch only what happens inside the browser window. It cannot see if a student is using a second phone to look up answers, nor can it detect if a student is using a separate computer or a hidden virtual machine to run other software. These are threats that exist outside the browser, and the researchers acknowledge that their tool cannot stop them. For those types of academic dishonesty, physical supervision or other security measures are still necessary. The study was also conducted in a controlled laboratory setting using automated scripts rather than real students taking a real exam. While this allowed for precise measurements, the researchers note that real-world conditions, with their messy internet connections and unpredictable human behavior, might produce slightly different results.

The true value of this work lies not in inventing a brand new way to catch academic dishonesty, but in bringing transparency to a technology that is already widely used but poorly understood. Many schools currently rely on commercial products that promise to solve the academic dishonesty problem, yet no independent, peer-reviewed study has ever measured how accurate or fast these products actually are. By building an open, documented version of this technology and testing it rigorously, the researchers have provided the first clear benchmark for the industry. They have shown that a lightweight browser extension can be highly accurate and fast without demanding heavy resources or invading student privacy. This evidence allows educational institutions to make decisions based on facts rather than marketing claims, offering a viable path forward for secure testing in an era where students bring their own devices to the classroom.