image copyrightNews1
“I committed cheating by taking test questions with a tiny camera and sending them to AI to check the answers on my smartphone.”
The confession of Mr. Han, who passed the tax accountant qualification test at the top of the list this year, is increasing the anxiety of test-takers ahead of the College Scholastic Ability Test. As cheating using artificial intelligence (AI) goes undetected even in front of supervisors, concerns are growing that something similar could happen in the CSAT.
Mr. A, who passed the 63rd tax accountant exam, appeared on a lawyer's YouTube channel on the 15th and admitted that he used a camera, smartphone, and generative AI in the first and second exams.
According to the Human Resources Development Service of Korea, Mr. A's seat was in the center of the second row based on the denomination during the first test, and in the center of the first row during the second test. He was not caught even in the front seat of the examination room where two supervisors were watching.
The discussion that was focused on identifying AI smart glasses is now expanding to how to respond to overall fraud that combines various devices and AI.
As the government strengthens its supervision measures ahead of the CSAT on the 19th, we looked at how far it can detect and prevent cheating using technology.
Government strengthens device return and considers blocking public Wi-Fi
Photo source: Attorney Lee Don-ho’s YouTube
This is not the first time that test malpractice using smart devices has been revealed.
In the national qualification test conducted by the Human Resources Development Service of Korea, there were six cases of detection of AI glasses this year. Last month, test takers were caught wearing or in possession of smart glasses at Yonsei University and Seoul National University's math essay test site. A survey by the Ministry of Education found that four cases of AI glasses-related fraud occurred in middle and high school grades this year.
End of TOP News
Accordingly, education authorities have deployed dedicated personnel to determine whether smart glasses are available at CSAT test centers across the country and have been providing device identification training to supervisors.
However, the question raised by this incident is not limited to how to distinguish specific devices. In a situation where AI can read test questions and provide answers, the question is whether the existing supervision system can respond to such cheating.
After Han's confession, there were a series of posts on the Korea Institute for Curriculum and Evaluation's website calling for effective measures against AI misconduct. There was also a request to postpone the CSAT until a solution is prepared.
The government held a meeting with relevant ministries on the 16th and began further response.
Prime Minister Han Seong-sook formed a government-wide emergency response team for the college entrance exam and ordered all smart devices to be returned and thoroughly checked for possession during the test. We also decided to actively consider blocking public Wi-Fi at the test site during CSAT time with the relevant organizations.
If you bring electronic devices that are prohibited during the CSAT, such as mobile phones, smart glasses, or smart watches, you must submit them to the supervisor before the start of the first period.
The government plans to check for signs of abnormalities even after the test and, if fraudulent activity is revealed, respond by invalidating the test, restricting test-taking qualifications, or requesting an investigation. We decided to receive reports through the reporting center and investigate suspected cases in cooperation with relevant agencies such as the police.
Regarding the overall national qualification test, the installation of smart search stations and the establishment of certified test locations are also being considered.
Response plans are largely divided into pre-inspection to prevent devices from entering the test site, measures to restrict communication and use during the test, and post-test to check for suspicious cases after the test.
Technological means include searches using radio detectors and metal detectors, and 'jamming', which blocks wireless communication by sending jamming waves. There are also ways to make camera filming and AI character recognition difficult by adding special patterns to the test paper or adjusting the lighting.
However, questions remain about how effectively these methods work in the actual test environment and what effect they have on test takers' problem solving and normal communication.
Can we stop it with just detection equipment?
Experts say that while technological means can help, installing the equipment alone cannot completely stop fraud.
A radio detector detects radio signals and confirms the presence of a device. However, the process of capturing the signal and then finding the actual user is not simple.
Lee Jae-seong, a professor in the AI department at Chung-Ang University, told the BBC, "To narrow down the range where signals occur within the test room, a person must carry a directional antenna and a mobile receiver," and pointed out that such a test could hinder concentration in situations where test takers are sensitive.
He went on to say, “Even if a suspicious signal is detected, whether or not a body search can actually be conducted has a completely different meaning,” adding that one must also be prepared for a situation where the test taker refuses the search or the search results turn up nothing.
Professor Lee proposed a plan to prevent devices from being brought in through entrance screening, but he also believed that search stations were not a perfect solution.
“The search bar is one of the alternatives currently being considered, but it cannot be perfect. Whether it exists or not must be judged by a person, and people can make mistakes.”
In the CSAT where many test takers enter at once, delays due to testing must be taken into consideration.
Professor Lee proposed a plan to conduct tests at the school entrance rather than installing a security checkpoint in each exam room, but said that securing additional personnel and equipment remains a challenge.
In addition, as each individual technology has weaknesses, it was necessary to review various measures together and increase the number of supervisors to reduce blind spots in the test room.
There are also cases where multiple stages of inspection are combined. Gaokao, China's university entrance exam, utilizes entrance search, communication blocking, and AI video analysis.
According to China's Hunan Provincial Board of Education and Examinations, this year, a 'double search' was conducted through a smart search bar and human inspection, and the wireless communication signal in the test room was blocked. During the test, AI detected abnormal behavior in the video and sent a warning image and a short video to supervisors. When AI finds a suspicious situation, a person checks it.
However, it is difficult to simply compare the domestic situation with China, which implements measures under a strong state-led administrative system. In Korea, expanding communication blocking or AI video surveillance requires social discussion and agreement surrounding test takers' rights, personal information protection, and related laws.
A gap that remains even if communication is cut off
Measures that restrict communication also have different effects depending on the scope of application. Even if you turn off public Wi-Fi, this does not mean that all wireless connections between mobile networks or devices are cut off.
There is also ‘on-device AI’ that answers questions within the smartphone without accessing an external server. These devices can operate even if external communications are blocked.
'Jamming' is mentioned as a more powerful method of blocking communication. This is a technology that makes it difficult for devices to send and receive information wirelessly by sending jamming waves to a specific frequency band. Although it can be effective against cheating using devices that are connected to the outside world in real time, it can also affect normal communication around the test site.
Joon-ho Hong, a professor of convergence security engineering at Sungshin Women's University, told the BBC, "There is a possibility that problems may occur, especially in emergency calls and safety-related communications. In Korea, regulatory issues under the Radio Wave Act must also be reviewed," adding, "There are realistic difficulties in uniformly introducing it to general test sites."
Methods of interfering with camera filming and AI character recognition by adding special patterns to the test paper or adjusting the lighting also need to be verified. Professor Hong explained that AI's image recognition ability is advancing and that special patterns or lighting can reduce test takers' readability or cause eye fatigue.
In addition to the technical effectiveness, the cost and preparation time required to secure and install the equipment must also be considered.
Seokbin Yoon, a special professor at Sogang University's AI and SW Graduate School, told the BBC that radio jamming devices are used to protect the president and key personnel, and to prevent terrorism, and said, "Whether this is done at all test centers is another question."
He pointed out that jamming can be effective, but the effectiveness relative to the input must be considered. The purpose is to judge effectiveness by considering the cost and time required to apply it to testing centers nationwide.
CSAT D-2, what needs to be improved immediately?
Professor Hong Joon-ho said that in situations where the CSAT is imminent, it is important to ensure that existing supervision procedures are working properly rather than rushing to bring in new equipment.
He said, “Effectiveness can be increased only when not only the performance of the equipment, but also detection procedures, supervisor proficiency, and confirmation procedures for suspicious devices are in place,” and explained that the management of imported goods, training of supervisors, and response procedures to suspicious situations must be strengthened.
This means that we must not only find the device, but also prepare who will check it and by what standards to make the final decision. In this process, standards are also needed to protect test takers who take the exam normally.
Professor Hong emphasized, “There is a need to establish objective verification standards and procedures to prevent unnecessary suspicions from test-takers who use regular glasses or medical assistive devices,” and added, “A balanced response is needed that ensures the fairness of the test while not interfering with the normal progress of the test.”
In the long term, it is pointed out that a sense of responsibility for the use of AI must be developed along with technology to assist supervision.
Professor Seokbin Yoon said that just as AI is used to respond to attacks using AI in the security field, there is a need to use AI in test supervision to complement the limitations of visual surveillance.
He emphasized that, in addition to cultivating AI utilization skills, education on ethics and social responsibility is also necessary. Beyond immediate responses to protect the fairness of tests, social discussions are needed on how to use AI and be responsible for the results.
He said, “As the AI era progresses, reliability will become a more important value.”




