In this blog post, we’ll examine the operating principles and safety features of self-driving cars, as well as the risks to cybersecurity, and explore just how far autonomous driving technology has actually advanced.
A rather special car has made an appearance at Seoul National University, where countless vehicles come and go every day. This is the story of SNUBer 2, which navigates the narrow, winding roads within the College of Engineering without a steering wheel—an essential component of driving. SNUBer 2 is Seoul National University’s autonomous vehicle that uses radar and moving object detection and tracking technology to perceive its surroundings and navigate. The Seoul National University Intelligent IT Research Center conducted research aimed at the practical application of autonomous vehicles through SNUBer 2. Compared to its predecessor, SNUBer 1, it was developed into a more advanced autonomous driving system through the application of new technologies, such as ultra-high-precision 3D maps. A high-precision 3D road map is an electronic map that represents road infrastructure and signage information in three dimensions. The College of Engineering at Seoul National University has also been steadily conducting research on sensor development, as well as algorithms and programming for vehicle control. This research has led to the testing and validation of autonomous driving technology on both the university campus and actual road environments.
The field of autonomous vehicles is one in which research is actively progressing alongside advancements in artificial intelligence. While the commercialization of autonomous vehicles was once considered a distant prospect, autonomous driving technology has now advanced to the point where it is being utilized in commercial services on public roads under limited conditions. For example, Waymo is currently operating fully autonomous, driverless services in several cities as of 2026 and is gradually expanding the scope of its autonomous driving systems. However, this does not mean that the technology has reached a level where it can drive safely on all roads and in all situations without human intervention. Furthermore, the psychological factor of not being able to control the vehicle directly can increase anxiety among consumers using self-driving cars. Several accidents related to self-driving technology have occurred in the past. In 2016, a fatal accident occurred involving a Tesla vehicle while its Autopilot was active, and in the same year, a Google self-driving test vehicle was also involved in an accident. In March 2018, an Uber self-driving test vehicle struck and killed a pedestrian in Tempe, Arizona. At the time, the vehicle was operating in autonomous mode, and the accident investigation identified several key causes, including the fact that the safety operator in the driver’s seat had failed to properly monitor the driving environment and the autonomous driving system, as well as deficiencies in Uber’s safety management system. It is true that skepticism regarding self-driving cars has spread due to these various accidents. However, the safety of self-driving cars should not be judged solely based on accident cases. One must understand both the operating principles and limitations of self-driving systems to distinguish between vague anxieties about the technology and actual risks.
For a self-driving car to operate, it must first perceive its surroundings using sensors such as cameras, radar, and the Global Positioning System (GPS). Once various data points are collected through these sensors, the data is processed digitally, and based on this analysis, the vehicle determines which direction to travel and at what speed. Furthermore, to discuss the safety of autonomous vehicles, it is necessary to understand their specific operating principles. Autonomous vehicles generally consist of several functional systems, including those for perceiving the vehicle’s surroundings, determining the vehicle’s location, and avoiding obstacles and determining the driving route. The vehicle is equipped with sensors such as cameras, radar, and LiDAR, as well as computers responsible for real-time control and computers that process video and sensor data; the information collected by each device is integrated and analyzed through the computing system. Autonomous vehicles operate by continuously repeating the process of recognizing the surrounding environment, determining their own position and situation, avoiding obstacles, and controlling their movement. These systems are designed to perform tasks ranging from driver-assistance features to advanced autonomous driving functions, including driving, braking, and responding to unexpected situations.
To maintain stability, autonomous vehicles can be designed with hardware sensors and software configured independently of one another or with distributed functionality, thereby minimizing the impact on overall system operation even if a problem occurs in a single device. For example, even if some cameras or laser sensors are damaged and fail to operate normally, the system is designed to continue driving or transition to a safe state by utilizing other sensors and systems. Additionally, the vehicle is designed with an emergency stop function to bring the vehicle to a halt when an unexpected situation arises, or to allow the vehicle control system to stop the vehicle directly if necessary. Each data point is transmitted to a computer via a network, and this data is extracted and utilized as needed by the respective algorithm modules. Therefore, functionally separating the algorithm modules from the sensor modules and configuring them independently can enhance the overall stability of the system even if errors occur in some sensors.
Other factors that can contribute to the instability of autonomous vehicles include hacking and cyberattacks. This is because the algorithms and data of autonomous vehicles are processed by artificial intelligence and computer programs. Since autonomous vehicles are connected to external networks and control the vehicle through various sensors and computer systems, cybersecurity is treated as a critical issue. In the past, public-key cryptographic technologies, such as RSA encryption, were widely used for network security. RSA encryption is based on fundamental number theory, particularly the computational difficulty of factorization. The core idea behind RSA is to make public a number created by multiplying two large prime numbers and to leverage the fact that it is difficult to factor that number back into its prime factors to ensure the security of the encryption. Mathematical research on prime numbers has a long history, and number theory—once considered a field of pure mathematics far removed from practical applications—is a prime example of how it has come to play a vital role in modern cryptography and information security. However, existing public-key cryptosystems, including RSA, may face new threats if future quantum computers advance sufficiently. In particular, Shor’s algorithm has the potential to undermine the security of RSA because it can efficiently factor integers on a sufficiently powerful quantum computer. Therefore, in fields where security is of the utmost importance—such as artificial intelligence and autonomous driving—simply increasing the time required for an attack is not sufficient; it is crucial to continuously advance cryptographic systems and prepare for new threats.
To address this, new encryption technologies are being researched to protect existing public-key cryptosystems from attacks by quantum computers. Prominent examples include quantum cryptography and quantum-resistant cryptography. Quantum cryptography is a method that utilizes the properties of quantum mechanics to detect eavesdropping during communication, while quantum-resistant cryptography refers to new cryptographic algorithms designed to withstand attacks using quantum computers. Quantum cryptography detects third-party interference by exploiting the quantum mechanical property that a quantum state can change during the measurement process. However, these technologies do not automatically solve all existing security issues; to ensure the security of actual autonomous driving systems, comprehensive protection is required not only for encryption but also for the vehicle’s internal network, software, communication networks, and update systems.
To date, it remains difficult to place complete trust in autonomous vehicles in all situations, both in terms of safety and technology. In particular, technical factors—such as mechanical failures or software errors—could potentially lead to accidents. Furthermore, as sensational accident reports receive extensive media coverage, the risks associated with autonomous driving technology may be exaggerated beyond reality or, conversely, viewed with excessive optimism.
From a technical standpoint, autonomous vehicles incorporate various sensors, control systems, and safety measures—such as redundancy—to reduce the mistakes that can occur during typical human driving and ensure safer operation. Additionally, research is ongoing to strengthen the security of vehicles and communication systems in response to concerns about hacking and cyberattacks. However, these safety measures do not completely eliminate the possibility of accidents. In fact, while autonomous driving technology is already being used in commercial services under limited conditions, we have not yet reached the stage where we can conclusively state that it is safer than human drivers in all situations. Therefore, what is important going forward is not to view autonomous vehicles as a perfect technology that never causes accidents, but rather to accurately understand their technical limitations and the potential for accidents, and to continuously develop safety and security systems to mitigate these risks.
