arXiv subjects

Dan Dongseong Kim

Publications and source records attributed to Dan Dongseong Kim.

12 recordsLinked to original sources

arXiv2026

Harnessing LLMs for Document-Guided Fuzzing of Python Libraries

Python libraries underpin deep learning, scientific computing, data analysis, and computer vision, making their reliability critical to downstream applications. Testing their APIs requires inputs that satisfy both per-parameter constraints and dependencies among parameters. Existing approaches either leave such constraints implicit in generated programs or rely on library-specific parsing rules. This paper introduces VistaFuzz, a document-guided fuzzing technique that uses a locally served open-sourced LLM to extract parameter specifications from API documents and generate inputs that satisfy both parameter constraints and inter-parameter dependencies. We evaluate VistaFuzz on 7,718 APIs across twelve Python libraries. Inter-parameter relationships occur in 40.1\% of tested APIs, and disabling their resolution reduces the valid generation rate on those APIs from above 95\% to 31.6\%--52.8\%. VistaFuzz reports 74 issues, of which 43 have been confirmed by developers and 29 have been fixed.

cs.SE↗

arXiv2026

Testing Deep Learning Library APIs via Cross-Framework Differential Fuzzing

Deep learning libraries underpin many safety- and reliability-critical applications, yet existing API-level testing techniques often rely on intra-library properties or CPU--GPU differential oracles and may miss defects that behave consistently across hardware backends. We present Xamt, a cross-framework differential fuzzing approach for deep learning library APIs. Xamt constructs and tests execution-validated groups of APIs intended to implement equivalent operations across seven libraries. It uses explicit API aliases and parameter-role normalization to construct candidate correspondences and validates them through pairwise execution and a group-level behavioral check on canonical ordinary inputs. The resulting groups are explored using variance-guided differential fuzzing with ordinary, boundary, and non-finite inputs. Crash and inconsistency oracles flag executions exhibiting abnormal termination or inconsistent outputs for subsequent reproduction and analysis. Across the seven libraries, Xamt constructs 676 execution-validated groups containing 2,563 matched APIs. Among these, Xamt identifies 72 independently reproduced discrepancy cases, including 4 crash cases and 68 output inconsistencies. Among the 72 developer reports, 25 have been confirmed, including 23 that have been fixed.

cs.SE↗

arXiv2026

Towards Effective Black-Box Adversarial Attacks on Deep Code Models via Structural and Identifier Perturbations

Deep code models (DCMs) are increasingly embedded in code intelligence tasks. However, their robustness under adversarial attacks remains insufficiently understood. Prior black-box attacks mainly rely on identifier- only substitutions or structural edits transferred from reference samples, yielding perturbation spaces defined largely independently of the attacked input. We introduce Strike, an input-conditioned black-box adversarial robustness-testing framework that constructs and searches a hierarchical perturbation space for each input. Strike first partitions code into blocks and uses an LLM to generate context-specific structural candidates, filtered for syntactic validity, ranked by similarity, and adaptively combined. It then refines the best structural variant through similarity-guided identifier substitutions drawn from dynamically constructed candidate pools. Evaluations across representative code intelligence tasks, including clone detection, vulnerability detection, and code summarization, demonstrate that Strike achieves higher attack success rates with competitive target- model query overhead. Strategy-specific static checks across all three tasks and execution-based validation on the executable Juliet vulnerability-detection subset provide complementary, task-bounded evidence regarding the validity of the generated variants. Representation similarity and human evaluation further indicate that the variants remain highly similar to the original code and contextually natural. Fine-tuning with Strike- generated samples improves cross-attack performance on fixed perturbed evaluation sets while preserving clean performance.

cs.CR↗

arXiv2026

APT-Agent: Automated Penetration Testing using Large Language Models

Penetration testing is essential to securing modern web infrastructures, yet traditional manual methods struggle to keep pace with their scale and complexity. Large Language Models (LLMs) offer new opportunities for automating these tasks, but existing approaches face two persistent challenges: hallucination of technical entities and insufficient long-term contextual memory. To address these issues, we present APT-Agent, a fully automated LLM-driven penetration testing framework that systematically orchestrates reconnaissance, exploitation, and exfiltration. APT-Agent introduces a hybrid rectification module to recover hallucinated commands and a command-specific memory architecture to preserve operational context across multi-step attack sequences. We evaluate our APT-Agent on Metasploitable 2 against seven vulnerable services spanning web, database, and network protocols. APT-Agent achieves an 84.29% end-to-end exploitation success rate, compared to 48.57% (Script Kiddie) and 18.57% (PentestGPT) under matched conditions. By reducing cognitive burden and minimizing reliance on human intervention, APT-Agent represents a step toward scalable, reliable, and cognitively efficient automation for penetration testing.

cs.CR↗

arXiv2026

DriveMind: A Dual Visual Language Model-based Reinforcement Learning Framework for Autonomous Driving

End-to-end autonomous driving systems map sensor data directly to control commands, but remain opaque, lack interpretability, and offer no formal safety guarantees. While recent vision-language-guided reinforcement learning (RL) methods introduce semantic feedback, they often rely on static prompts and fixed objectives, limiting adaptability to dynamic driving scenes. We present DriveMind, a unified semantic reward framework that integrates: (i) a contrastive Vision-Language Model (VLM) encoder for stepwise semantic anchoring; (ii) a novelty-triggered VLM encoder-decoder, fine-tuned via chain-of-thought (CoT) distillation, for dynamic prompt generation upon semantic drift; (iii) a hierarchical safety module enforcing kinematic constraints (e.g., speed, lane centering, stability); and (iv) a compact predictive world model to reward alignment with anticipated ideal states. DriveMind achieves 19.4 +/- 2.3 km/h average speed, 0.98 +/- 0.03 route completion, and near-zero collisions in CARLA Town 2, outperforming baselines by over 4% in success rate. Its semantic reward generalizes zero-shot to real dash-cam data with minimal distributional shift, demonstrating robust cross-domain alignment and potential for real-world deployment.

cs.RO↗

arXiv2026

Risk-Aware Human-in-the-Loop Framework with Adaptive Intrusion Response for Autonomous Vehicles

Autonomous vehicles must remain safe and effective when encountering rare long-tailed scenarios or cyber-physical intrusions during driving. We present RAIL, a risk-aware human-in-the-loop framework that turns heterogeneous runtime signals into calibrated control adaptations and focused learning. RAIL fuses three cues (curvature actuation integrity, time-to-collision proximity, and observation-shift consistency) into an Intrusion Risk Score (IRS) via a weighted Noisy-OR. When IRS exceeds a threshold, actions are blended with a cue-specific shield using a learned authority, while human override remains available; when risk is low, the nominal policy executes. A contextual bandit arbitrates among shields based on the cue vector, improving mitigation choices online. RAIL couples Soft Actor-Critic (SAC) with risk-prioritized replay and dual rewards so that takeovers and near misses steer learning while nominal behavior remains covered. On MetaDrive, RAIL achieves a Test Return (TR) of 360.65, a Test Success Rate (TSR) of 0.85, a Test Safety Violation (TSV) of 0.75, and a Disturbance Rate (DR) of 0.0027, while logging only 29.07 training safety violations, outperforming RL, safe RL, offline/imitation learning, and prior HITL baselines. Under Controller Area Network (CAN) injection and LiDAR spoofing attacks, it improves Success Rate (SR) to 0.68 and 0.80, lowers the Disengagement Rate under Attack (DRA) to 0.37 and 0.03, and reduces the Attack Success Rate (ASR) to 0.34 and 0.11. In CARLA, RAIL attains a TR of 1609.70 and TSR of 0.41 with only 8000 steps.

cs.AI↗

arXiv2025

DASH: Deception-Augmented Shared Mental Model for a Human-Machine Teaming System

We present DASH (Deception-Augmented Shared mental model for Human-machine teaming), a novel framework that enhances mission resilience by embedding proactive deception into Shared Mental Models (SMM). Designed for mission-critical applications such as surveillance and rescue, DASH introduces "bait tasks" to detect insider threats, e.g., compromised Unmanned Ground Vehicles (UGVs), AI agents, or human analysts, before they degrade team performance. Upon detection, tailored recovery mechanisms are activated, including UGV system reinstallation, AI model retraining, or human analyst replacement. In contrast to existing SMM approaches that neglect insider risks, DASH improves both coordination and security. Empirical evaluations across four schemes (DASH, SMM-only, no-SMM, and baseline) show that DASH sustains approximately 80% mission success under high attack rates, eight times higher than the baseline. This work contributes a practical human-AI teaming framework grounded in shared mental models, a deception-based strategy for insider threat detection, and empirical evidence of enhanced robustness under adversarial conditions. DASH establishes a foundation for secure, adaptive human-machine teaming in contested environments.

cs.HC↗

arXiv2025

XAMT: Cross-Framework API Matching for Testing Deep Learning Libraries

Deep learning powers critical applications such as autonomous driving, healthcare, and finance, where the correctness of underlying libraries is essential. Bugs in widely used deep learning APIs can propagate to downstream systems, causing serious consequences. While existing fuzzing techniques detect bugs through intra-framework testing across hardware backends (CPU vs. GPU), they may miss bugs that manifest identically across backends and thus escape detection under these strategies. To address this problem, we propose XAMT, a cross-framework fuzzing method that tests deep learning libraries by matching and comparing functionally equivalent APIs across different frameworks. XAMT matches APIs using similarity-based rules based on names, descriptions, and parameter structures. It then aligns inputs and applies variance-guided differential testing to detect bugs. We evaluated XAMT on five popular frameworks, including PyTorch, TensorFlow, Keras, Chainer, and JAX. XAMT matched 839 APIs and identified 238 matched API groups, and detected 17 bugs, 12 of which have been confirmed. Our results show that XAMT uncovers bugs undetectable by intra-framework testing, especially those that manifest consistently across backends. XAMT offers a complementary approach to existing methods and offers a new perspective on the testing of deep learning libraries.

cs.SE↗

arXiv2025

Harnessing LLMs for Document-Guided Fuzzing of OpenCV Library

The combination of computer vision and artificial intelligence is fundamentally transforming a broad spectrum of industries by enabling machines to interpret and act upon visual data with high levels of accuracy. As the biggest and by far the most popular open-source computer vision library, OpenCV library provides an extensive suite of programming functions supporting real-time computer vision. Bugs in the OpenCV library can affect the downstream computer vision applications, and it is critical to ensure the reliability of the OpenCV library. This paper introduces VISTAFUZZ, a novel technique for harnessing large language models (LLMs) for document-guided fuzzing of the OpenCV library. VISTAFUZZ utilizes LLMs to parse API documentation and obtain standardized API information. Based on this standardized information, VISTAFUZZ extracts constraints on individual input parameters and dependencies between these. Using these constraints and dependencies, VISTAFUZZ then generates new input values to systematically test each target API. We evaluate the effectiveness of VISTAFUZZ in testing 330 APIs in the OpenCV library, and the results show that VISTAFUZZ detected 17 new bugs, where 10 bugs have been confirmed, and 5 of these have been fixed.

cs.SE↗

arXiv2025

Advancing Human-Machine Teaming: Concepts, Challenges, and Applications

Human-Machine Teaming (HMT) is revolutionizing collaboration across domains such as defense, healthcare, and autonomous systems by integrating AI-driven decision-making, trust calibration, and adaptive teaming. This survey presents a comprehensive taxonomy of HMT, analyzing theoretical models, including reinforcement learning, instance-based learning, and interdependence theory, alongside interdisciplinary methodologies. Unlike prior reviews, we examine team cognition, ethical AI, multi-modal interactions, and real-world evaluation frameworks. Key challenges include explainability, role allocation, and scalable benchmarking. We propose future research in cross-domain adaptation, trust-aware AI, and standardized testbeds. By bridging computational and social sciences, this work lays a foundation for resilient, ethical, and scalable HMT systems.

cs.HC↗

arXiv2024

MTDSense: AI-Based Fingerprinting of Moving Target Defense Techniques in Software-Defined Networking

Moving target defenses (MTD) are proactive security techniques that enhance network security by confusing the attacker and limiting their attack window. MTDs have been shown to have significant benefits when evaluated against traditional network attacks, most of which are automated and untargeted. However, little has been done to address an attacker who is aware the network uses an MTD. In this work, we propose a novel approach named MTDSense, which can determine when the MTD has been triggered using the footprints the MTD operation leaves in the network traffic. MTDSense uses unsupervised clustering to identify traffic following an MTD trigger and extract the MTD interval. An attacker can use this information to maximize their attack window and tailor their attacks, which has been shown to significantly reduce the effectiveness of MTD. Through analyzing the attacker's approach, we propose and evaluate two new MTD update algorithms that aim to reduce the information leaked into the network by the MTD. We present an extensive experimental evaluation by creating, to our knowledge, the first dataset of the operation of an IP-shuffling MTD in a software-defined network. Our work reveals that despite previous results showing the effectiveness of MTD as a defense, traditional implementations of MTD are highly susceptible to a targeted attacker.

cs.CR↗

arXiv2022

Liuer Mihou: A Practical Framework for Generating and Evaluating Grey-box Adversarial Attacks against NIDS

Due to its high expressiveness and speed, Deep Learning (DL) has become an increasingly popular choice as the detection algorithm for Network-based Intrusion Detection Systems (NIDSes). Unfortunately, DL algorithms are vulnerable to adversarial examples that inject imperceptible modifications to the input and cause the DL algorithm to misclassify the input. Existing adversarial attacks in the NIDS domain often manipulate the traffic features directly, which hold no practical significance because traffic features cannot be replayed in a real network. It remains a research challenge to generate practical and evasive adversarial attacks. This paper presents the Liuer Mihou attack that generates practical and replayable adversarial network packets that can bypass anomaly-based NIDS deployed in the Internet of Things (IoT) networks. The core idea behind Liuer Mihou is to exploit adversarial transferability and generate adversarial packets on a surrogate NIDS constrained by predefined mutation operations to ensure practicality. We objectively analyse the evasiveness of Liuer Mihou against four ML-based algorithms (LOF, OCSVM, RRCF, and SOM) and the state-of-the-art NIDS, Kitsune. From the results of our experiment, we gain valuable insights into necessary conditions on the adversarial transferability of anomaly detection algorithms. Going beyond a theoretical setting, we replay the adversarial attack in a real IoT testbed to examine the practicality of Liuer Mihou. Furthermore, we demonstrate that existing feature-level adversarial defence cannot defend against Liuer Mihou and constructively criticise the limitations of feature-level adversarial defences.

cs.CR↗