arXiv subjects

Min Geun Song

Publications and source records attributed to Min Geun Song.

4 recordsLinked to original sources

arXiv2025

Autoencoder-based Denoising Defense against Adversarial Attacks on Object Detection

Deep learning-based object detection models play a critical role in real-world applications such as autonomous driving and security surveillance systems, yet they remain vulnerable to adversarial examples. In this work, we propose an autoencoder-based denoising defense to recover object detection performance degraded by adversarial perturbations. We conduct adversarial attacks using Perlin noise on vehicle-related images from the COCO dataset, apply a single-layer convolutional autoencoder to remove the perturbations, and evaluate detection performance using YOLOv5. Our experiments demonstrate that adversarial attacks reduce bbox mAP from 0.2890 to 0.1640, representing a 43.3% performance degradation. After applying the proposed autoencoder defense, bbox mAP improves to 0.1700 (3.7% recovery) and bbox mAP@50 increases from 0.2780 to 0.3080 (10.8% improvement). These results indicate that autoencoder-based denoising can provide partial defense against adversarial attacks without requiring model retraining.

cs.CR↗

arXiv2024

RTPS Attack Dataset Description

This paper explains all about our RTPS datasets. We collect malicious/benign packet data by injecting attack data in an Unmanned Ground Vehicle (UGV) in the normal state. We assembled the testbed, consisting of UGV, Controller, PC, and Router. We collect this dataset in the UGV part of our testbed. We conducted two types of attack "Command Injection" and "Command Injection with ARP Spoofing" on our testbed. The data collection time is 180, 300, 600, and 1200. The scenario has 30 each on collection time, 240 total. We expect this dataset to contribute to the development of defense technologies like anomaly detection to address security threat issues in ROS2 networks and Fast-DDS implements.

cs.CR↗

arXiv2023

C-ITS Environment Modeling and Attack Modeling

As technology advances, cities are evolving into smart cities, with the ability to process large amounts of data and the increasing complexity and diversification of various elements within urban areas. Among the core systems of a smart city is the Cooperative-Intelligent Transport Systems (C-ITS). C-ITS is a system where vehicles provide real-time information to drivers about surrounding traffic conditions, sudden stops, falling objects, and other accident risks through roadside base stations. It consists of road infrastructure, C-ITS centers, and vehicle terminals. However, as smart cities integrate many elements through networks and electronic control, they are susceptible to cybersecurity issues. In the case of cybersecurity problems in C-ITS, there is a significant risk of safety issues arising. This technical document aims to model the C-ITS environment and the services it provides, with the purpose of identifying the attack surface where security incidents could occur in a smart city environment. Subsequently, based on the identified attack surface, the document aims to construct attack scenarios and their respective stages. The document provides a description of the concept of C-ITS, followed by the description of the C-ITS environment model, service model, and attack scenario model defined by us.

cs.CR↗

arXiv2023

AI-based Attack Graph Generation

With the advancement of IoT technology, many electronic devices are interconnected through networks, communicating with each other and performing specific roles. However, as numerous devices join networks, the threat of cyberattacks also escalates. Preventing and detecting cyber threats are crucial, and one method of preventing such threats involves using attack graphs. Attack graphs are widely used to assess security threats within networks. However, a drawback emerges as the network scales, as generating attack graphs becomes time-consuming. To overcome this limitation, artificial intelligence models can be employed. By utilizing AI models, attack graphs can be created within a short period, approximating optimal outcomes. AI models designed for attack graph generation consist of encoders and decoders, trained using reinforcement learning algorithms. After training the AI models, we confirmed the model's learning effectiveness by observing changes in loss and reward values. Additionally, we compared attack graphs generated by the AI model with those created through conventional methods.

cs.CR↗