Background Android is the most widely used operating system all over the world. Due to its open nature, the Android operating system has become the target of malicious coders. Ensuring privacy and security is of great importance to Android users. Methods In this study, a hybrid architecture is proposed for the detection of Android malware from the permission information of applications. The proposed architecture combines the feature extraction power of the convolutional neural network (CNN) architecture and the decision making capability of fuzzy logic. Our method extracts features from permission information with a small number of filters and convolutional layers, and also makes the feature size suitable for ANFIS input. In addition, it allows the permission information to affect the classification without being neglected. In the study, malware was obtained from two different sources and two different data sets were created. In the first dataset, Drebin was used for malware applications, and in the second dataset, CICMalDroid 2020 dataset was used for malware applications. For benign applications, the Google Play Store environment was used. Results With the proposed method, 92% accuracy in the first data set and 92% F-score value in the weighted average was achieved. In the second data set, an accuracy of 94.6% and an F-score of 94.6% on the weighted average were achieved. The results obtained in the study show that the proposed method outperforms both classical machine learning algorithms and fuzzy logic-based studies.
Bu çalışmada, son yıllarda görüntü sınıflandırmada artan oranda ilgi gören derin öğrenme ve görüntü işleme yöntemleri kullanılarak kötü huylu (malignant) cilt lezyonlarının erken teşhisini kolaylaştırıcı yapay zekâ tabanlı sınıflandırma deneyleri gerçekleştirilmiştir. Melanom, en kötü huylu ve az görülen bir kanser türü olduğundan dolayı derin öğrenme mimarisini eğitmek için yeterli sayıda eğitim ve test görüntüsü bulmak zordur. Bu nedenle artırılmış veri seti oluşturulmuş ve 6 farklı derin öğrenme mimarisi ile eğitim yapılmıştır. Kötü huylu ve iyi huylu cilt lezyonlarını sınıflandırmak için popüler olan AlexNet, DenseNet-121, ResNet-18, ResNet-34, SqueezeNet ve VGGNet-16 mimarileri kullanılmıştır. Deneyler HAM10000 veri seti üzerinde artırma yapılarak gerçekleştirilmiştir. Deneyler sonucunda en başarılı sonuçları veren Resnet-34 mimarisi ile ortalama %87,5 doğruluk oranı, %94 AUC skoru, %84,5 F-skoru, %87,6 kesinlik değeri elde edilmiştir. Diğer derin öğrenme mimarilerinden elde edilen sonuçlar ve karşılaştırmalı analizler de çalışmada ayrıca sunulmuştur.
The growing trend toward vehicles being connected to various unidentified devices, such as other vehicles or infrastructure, increases the possibility of external attacks on“vehicle cybersecurity (VC). Detection of intrusion is a very important part of network security for vehicles such as connected vehicles, that have open connectivity, and self-driving vehicles. Consequently, security has become an important requirement in trying to protect these vehicles as attackers have become more sophisticated in using malware that can penetrate and harm vehicle control units as technology advances. Thus, ensuring the vehicles and the network are safe is very important for the growth of the automotive industry and for people to have more faith in it. In this study, a machine learning-based detection approach using hybrid analysis-based particle swarm optimization (PSO) and an adaptive genetic algorithm (AGA) is presented for Android malware detection in auto-driving vehicles. The “CCCS-CIC-AndMal-2020” dataset containing 13 different malware categories and 9504 hybrid features was used for the experiments. In the proposed approach, firstly, feature selection is performed by applying PSO to the features in the dataset. In the next step, the performance of XGBoost and random forest (RF) machine learning classifiers is optimized using the AGA. In the experiments performed, a 99.82% accuracy and F-score were obtained with the XGBoost classifier, which was developed using PSO-based feature selection and AGA-based hyperparameter optimization. With the random forest classifier, a 98.72% accuracy and F-score were achieved. Our results show that the application of PSO and an AGA greatly increases the performance in the classification of the information obtained from the hybrid analysis.
Diabetic retinopathy is an eye disease that occurs with damage to the retina and has many different complications, ranging from permanent blindness. The aim of this study is to develop a (convolutional neural network) CNN model that determines with high accuracy whether fundus images are diabetic retinopathy. The performance of the model has been verified in Kaggle APTOS 2019 dataset with AlexNET and VggNET-16 deep transfer learning algorithms. Various image processing techniques have been used as well as deep learning methods to further improve the classification performance. Images in the data set were rescaled to 224 × 224 × 3 and converted to Grayscale color space. Besides Gauss filter applied to eliminate the noise in the images. The area under the curve (AUC), precision, recall, and accuracy metrics of the deep transfer learning models used in this study were compared. The AlexNet model achieved a 98.6% AUC score, 95.2% accuracy, and the VggNET-16 model achieved a 99.6% AUC score and 98.1% accuracy. VggNET-16 was found to have higher confidence. Our results show that with the correct optimization of the CNN model applied in diabetic retinopathy classification, deep transfer learning models can achieve high performance and can be used in the detection of diabetic retinopathy patients.
Hazelnut is a product with high nutritional and economic value. In maintaining the quality value of hazelnut, the classification process is of great importance. In the present day, the quality classification of hazelnuts and other crops is performed in general manually, and so it is difficult and costly. Performing this classification with modern agricultural techniques is much more important in terms of quality. This study was based on a model intended to detect hazelnut quality. The model is about the establishment of an artificial intelligence-based classification system that can detect the hidden defects of hazelnuts. In the developed model, the visuals used in the dataset are divided into training and test groups. In the model, hazelnuts are divided into 5 classes according to their quality using AlexNet architecture and modern deep learning (DL) techniques instead of traditional hazelnut classification methods. In this model developed based on artificial intelligence, a very good approach was presented with the accurate classification of 99%. Moreover, the values regarding precision and recall were also determined at 98.7% and 99.6%, respectively. This study is important in terms of becoming widespread information technology use and computer-assisted applications in the agricultural economics field such as product classification, quality, and control.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.