arXiv subjects

Yonghwi Kwon

Publications and source records attributed to Yonghwi Kwon.

4 recordsLinked to original sources

arXiv2026

PatchBench: Evaluating AI Agents for Vulnerability Patching

AI agents have recently demonstrated strong performance in automated vulnerability patching. However, existing evaluations often validate a patch only by testing whether the provided Proof-of-Concept (PoC) input still triggers a crash. This leaves two key threats to validity: agents may reproduce memorized historical developer patches, or they may generate surface-level fixes that only suppress the reported crash. We study these concerns for C/C++ vulnerability patching. We introduce a patch similarity metric to detect memorized patches. On average, 25% of the agent patches exhibit substantial similarity to historical developer patches, indicating that patch memorization is a real threat to the validity of vulnerability patching evaluations. Meanwhile, agents also frequently exploit benchmark structures to pass patch validation by patching on the crash stack trace to suppress the crash, rather than localizing and fixing the root cause of the vulnerabilities. To handle these issues, we propose PatchBench, a new benchmark for evaluating AI agents on realistic vulnerability patching tasks. PatchBench selects vulnerabilities whose ground-truth fixes lie outside the crash stack and uses vulnerability transplant and code mutations to migrate historical vulnerabilities into new repository contexts, reducing the risks of surface-level fixes and patch memorization. We develop new patch validation methods that thoroughly evaluate both security and semantic correctness of agent patches. Across 11 state-of-the-art agents, including the top three AIxCC agents, the original PoC-only validation inflates the patching task solve rate of agents by 1.83$\times$ on average. Our results reveal key limitations of current patching agents and point to future research directions for more reliable vulnerability repair.

cs.CR↗

arXiv2022

DriveFuzz: Discovering Autonomous Driving Bugs through Driving Quality-Guided Fuzzing

Autonomous driving has become real; semi-autonomous driving vehicles in an affordable price range are already on the streets, and major automotive vendors are actively developing full self-driving systems to deploy them in this decade. Before rolling the products out to the end-users, it is critical to test and ensure the safety of the autonomous driving systems, consisting of multiple layers intertwined in a complicated way. However, while safety-critical bugs may exist in any layer and even across layers, relatively little attention has been given to testing the entire driving system across all the layers. Prior work mainly focuses on white-box testing of individual layers and preventing attacks on each layer. In this paper, we aim at holistic testing of autonomous driving systems that have a whole stack of layers integrated in their entirety. Instead of looking into the individual layers, we focus on the vehicle states that the system continuously changes in the driving environment. This allows us to design DriveFuzz, a new systematic fuzzing framework that can uncover potential vulnerabilities regardless of their locations. DriveFuzz automatically generates and mutates driving scenarios based on diverse factors leveraging a high-fidelity driving simulator. We build novel driving test oracles based on the real-world traffic rules to detect safety-critical misbehaviors, and guide the fuzzer towards such misbehaviors through driving quality metrics referring to the physical states of the vehicle. DriveFuzz has discovered 30 new bugs in various layers of two autonomous driving systems (Autoware and CARLA Behavior Agent) and three additional bugs in the CARLA simulator. We further analyze the impact of these bugs and how an adversary may exploit them as security vulnerabilities to cause critical accidents in the real world.

cs.RO↗

arXiv2022

Non-Stop & Non-Breakable Code Review Services in a Distributed System: Detecting Issues in Real Time

The two most significant bottlenecks in code merging are the build process and the unit tests. However, as the number of items to be checked in a code review increases, that code review becomes a bottleneck for code merging as well. Because of the dependency structure between code review services, an error in one service affects the entire service. As a result, whenever a service error occurs, it is crucial to have methods for determining which code review service has ultimately caused the error. With the goal of achieving a non-stop & non-breakable code review service, this paper describes an early error detection method along with a case study of the service.

cs.SE↗

arXiv2021

Spinner: Automated Dynamic Command Subsystem Perturbation

Injection attacks have been a major threat to web applications. Despite the significant effort in thwarting injection attacks, protection against injection attacks remains challenging due to the sophisticated attacks that exploit the existing protection techniques' design and implementation flaws. In this paper, we develop Spinner, a system that provides general protection against input injection attacks, including OS/shell command, SQL, and XXE injection. Instead of focusing on detecting malicious inputs, Spinner constantly randomizes underlying subsystems so that injected inputs (e.g., commands or SQL queries) that are not properly randomized will not be executed, hence prevented. We revisit the design and implementation choices of previous randomization-based techniques and develop a more robust and practical protection against various sophisticated input injection attacks. To handle complex real-world applications, we develop a bidirectional analysis that combines forward and backward static analysis techniques to identify intended commands or SQL queries to ensure the correct execution of the randomized target program. We implement Spinner for the shell command processor and two different database engines (MySQL and SQLite) and in diverse programming languages including C/C++, PHP, JavaScript and Lua. Our evaluation results on 42 real-world applications including 27 vulnerable ones show that it effectively prevents a variety of input injection attacks with low runtime overhead (around 5%).

cs.CR↗