What if someone could secretly hear the answers through a tiny wireless earbud during an online test? Preventing exactly that is the aim of a new "ear scanning" technology.

The Duolingo English Test (DET), accepted by over 6,000 educational institutions worldwide, is rolling out mandatory ear scanning for all test-takers. Using machine learning to detect hidden wireless earbuds, it points to a new standard in online exam security. Here's how it works.

Why "Ear Scanning" Became Necessary

As online testing has become more common, cheating methods have grown more sophisticated. One major concern is the use of wireless earbuds. Modern earbuds have shrunk so much that they are virtually invisible to the naked eye. Test-takers have been caught hiding tiny devices deep in their ear canals to receive answers from accomplices outside the room.

In Japan, the 2022 Common University Entrance Test saw a major scandal when questions were leaked via smartphone during the exam. Korean universities have also reported group cheating through phone calls and social media during online exams. For test administrators worldwide, preventing this kind of fraud has become an urgent priority.

How Duolingo's "Ear Scan" Works

The "Ear Scan" that DET has introduced requires all test-takers to scan both ears with their smartphone camera before the exam begins.

Test-takers first point the camera at one ear from a short distance, then bring the same ear closer for a second pass. This far-and-near scan is repeated for both ears, and the whole process takes about one to two minutes.

During those minutes, machine learning algorithms check whether any earbud or audio-receiving device is hidden inside the ear. The system works across all smartphone models, keeping security consistent regardless of the device.

Ear Biometrics: As Unique as Fingerprints

Like fingerprints, the shape of the ear varies from person to person, and it has long been studied as a biometric identifier.

In 2016, NEC and Nagaoka University of Technology announced "ear acoustic authentication." A microphone-equipped earphone sends sound toward the ear canal, and the reflected sound is used to identify the individual, reportedly reaching over 99% accuracy in about one second. The shape of the ear canal differs from person to person, with distinguishable differences even between identical twins.

The ear's shape also stays largely unchanged from around age 8 to 70, allowing stable authentication over long periods. Since the COVID-19 pandemic made facial recognition harder because of masks, ear biometrics has drawn attention as a next-generation identification method.

That said, DET's Ear Scan is not built for identity verification. Its sole purpose is to detect, from the camera image, whether a device is hidden inside the ear.

Multi-Layered Security for Online Testing

Beyond Ear Scan, the DET layers several other measures into a comprehensive defense.

First, there is a dual AI-and-human proctoring system. AI analyzes the test video to flag suspicious behavior, and human proctors then review whatever gets flagged. Eye movement and unusual sounds are among the signals it watches for.

Second, the test is adaptive. Drawing on a database of tens of thousands of questions, it adjusts what appears in real time based on each response, so no two tests are alike. That sharply reduces the risk of leaked questions or pre-planned cheating.

Third, a required secondary camera and a 360-degree room scan monitor the whole testing environment. Test-takers must position their camera so that their keyboard, screen, and upper body are visible.

Adoption in Japan and Impact on University Admissions

In Japan, a number of universities and institutions now accept DET scores as proof of English proficiency. Worldwide, more than 90% of the top 100 universities in the U.S. News ranking accept it, making DET a growing third option alongside TOEFL and IELTS.

What sets DET apart is its accessibility: a $70 fee (less than half that of TOEFL or IELTS), around-the-clock at-home testing, and results within 48 hours, conveniences that traditional English tests can't match.

Japan's Ministry of Education has taken notice too. In November 2025, its working group on comprehensive English assessment for university admissions discussed anti-cheating measures across the various testing bodies, a sign that online-test integrity is becoming a national policy concern.

The Technology-vs-Cheating Arms Race

Ear Scan is one example of how security has to keep pace with technology. Duolingo says it is also working on detection for deepfakes and generative-AI-assisted cheating.

Online testing widens fair access by removing geographic and economic barriers. The security that protects that fairness has to evolve just as fast.

In Japan, there's ongoing debate about preventing cheating in online exams. What about in your country? How are online tests and biometric security measures handled where you live? We'd love to hear about it.

References