Earticle

본 연구는 실시간 객체 탐지 기술인 YOLO에 대해 YOLOv8부터 YOLO12까지의 버전과 모델에 대해 소규모 커스텀 데 이터셋 환경에서 성능평가를 수행하여, 제한된 데이터와 자원을 보유한 실제 응용 환경에서의 모델 선택 가이드라인을 제시한 다. 교통 환경 기반 6개 클래스, 2,664장, 총 6,354개의 라벨링 커스텀 데이터셋을 구축하고, Google Colaboratory(NVIDIA Tesla T4 GPU) 환경에서 동일한 학습 조건으로 YOLOv8부터 YOLO12까지 총 26개 모델을 학습·평가하였다. 연구 결과, YOLOv9m이 mAP50-95 0.520으로 자체평가 기준 중 최고 성능을 기록하였으며, 대부분의 버전에서 소형(s) 또는 중형(m) 모 델이 시간 대비 뛰어난 성능을 보였다. 대형(l) 및 초대형(x) 모델은 학습 시간이 2~3배 증가했음에도 성능은 정체되거나 감소하 였다. 공식 벤치마크와 달리 소규모 데이터셋에서는 중간 크기 이상 모델에서 성능이 감소하는 비선형적 패턴이 나타났으며, YOLO 버전보다 모델 크기 선택이 성능에 더 큰 영향을 미치는 것으로 확인되었다. 본 연구는 제한된 데이터 환경에서의 실용적 버전 및 모델 선택 기준을 제시하며, 향후 하이퍼파라미터 최적화 및 도메인이나 데이터 규모에 따른 비교 연구가 필요하다.

This study evaluates real-time object detection models in the YOLO series, from YOLOv8 to YOLO12, using a small-scale custom dataset and proposes practical guidelines for model selection in resource-constrained environments. A custom traffic-based dataset consisting of 2,664 images, 6 classes, and a total of 6,354 labeled objects was constructed. All experiments were conducted under identical training conditions using Google Colaboratory with an NVIDIA Tesla T4 GPU, and a total of 26 models from YOLOv8 to YOLO12 were trained and evaluated. The results show that YOLOv9m achieved the highest performance among the evaluated models, with an mAP50–95 of 0.520. In most versions, small (s) and medium (m) models demonstrated superior performance relative to training time. Large (l) and extra-large (x) models showed required 2–3 times longer training, while their performance remained stagnant or decreased. Unlike official benchmarks, a nonlinear performance pattern was observed in the small-scale dataset, where performance decreased in models of medium size or larger. Furthermore, model size had a greater impact on performance than the YOLO version itself. This study provides practical guidelines for selecting appropriate YOLO versions and model sizes in resource-constrained environments. Future work should explore hyperparameter optimization and extend comparative analyses across different domains and dataset scales.

Ransomware is one of the most significant cybersecurity threats facing the world. In this research we designed and constructed a custom cybersecurity AI dataset for ransomware detection. We then evaluated the dataset using different machine learning models. The dataset was constructed using Cuckoo Sandbox where raw ransomware samples were analyzed to extract key features such as API calls, DLL usage, file operations, network activity, process creation and registry changes. These were then carefully labeled as either ransomware or benign. For evaluation purposes, the custom cybersecurity AI dataset was utilized to train and test various machine learning models. The dataset was split into 80% for training and 20% for testing. Logistic Regression, Random Forest, K-Nearest Neighbors (KNN), and XGBoost models were used to evaluate the resulting custom Cybersecurity AI Dataset. We obtained higher results of accuracy, precision, recall, and F1 scores evaluation metrics. Moreover, our results demonstrate the robustness of a combination of well-designed custom Cybersecurity AI Datasets and machine learning techniques in enhancing ransomware detection mechanisms as well as providing a framework for future cybersecurity applications

As cyber threats continue to evolve in sophistication, there is urgent need for intelligent, adaptive and context aware intrusion detection systems. In this paper, we present an intrusion detection framework that employs deep learning models to detect anomalies in network traffic using custom dataset. The dataset was constructed in a controlled lab environment using various intrusion attack scenarios such as DoS, SSH abuse and VPN exploitation. Deep learning models were then applied to detect the intrusions. The performance of the models in performing detection tasks were evaluated using metrics of accuracy, precision, recall and F1- score. The results that were obtained indicate that hybrid model achieved the best results with overall accuracy of 0.99 followed by transformer (0.98) and LSTM model (0.97) being the last. This study highlights the potential of leveraging well designed custom IDS datasets and deep learning techniques to enhance intrusion detection mechanisms thereby providing a robust framework for intrusion detection applications.

In the current fast changing cyberspace, ransomware has continued to be a formidable threat. Through this research, using deep reinforcement learning and adversarial attack models, we undertook performance analysis evaluation of a locally constructed ransomware dataset. The dataset contained key dynamic features that were extracted from raw ransomware samples processed in Cuckoo sandbox environment. Our approach combined supervised learning for initial detection and Deep Q-Network (DQN) algorithm for adaptive behavioral analysis. An L2 Projected Gradient Descent (L2PGD) adversarial attack was then carried out to evaluate the robustness of both security and stability of the ransomware detection model. The results that were obtained demonstrated that Deep Reinforcement Learning (DRL) can effectively classify samples as benign and ransomware. Moreover, the successful adversarial attack underscores the need for improved robustness measures in artificial intelligence models.