Malicious Software Classification Using Transfer Learning of ResNet-50 Deep Neural Network

Rezende, Edmar; Ruppert, Guilherme C. S.; Carvalho, Tiago; Ramos, Fábio; Geus, Paulo Lício de

doi:10.1109/icmla.2017.00-19

Cited by 185 publications

(106 citation statements)

References 8 publications

Supporting

Mentioning

103

Contrasting

Unclassified

Order By: Relevance

“…Anomaly detection is performed separately on the 1D feature vectors and on the grayscale images. We described image classification approaches which are state-of-the-art for malware detection [25][26][27][28]. This motivated us to adopt image classification for insider threat detection.…”

Section: Proposed Methodsmentioning

confidence: 99%

“…There are existing works that have effectively applied image classification approaches and deep learning models for malware detection in the domain of cybersecurity. Transfer learning has been applied to malware classification [25] with an accuracy of 98.65%. Kancherla and Mukkamala [26] proposed malware detection by representing the machine code as grayscale images and extracting intensity-based and texture-based features.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Image-Based Feature Representation for Insider Threat Classification

2020

View full text Add to dashboard Cite

Cybersecurity attacks can arise from internal and external sources. The attacks perpetrated by internal sources are also referred to as insider threats. These are a cause of serious concern to organizations because of the significant damage that can be inflicted by malicious insiders. In this paper, we propose an approach for insider threat classification which is motivated by the effectiveness of pre-trained deep convolutional neural networks (DCNNs) for image classification. In the proposed approach, we extract features from usage patterns of insiders and represent these features as images. Hence, images are used to represent the resource access patterns of the employees within an organization. After construction of images, we use pre-trained DCNNs for anomaly detection, with the aim to identify malicious insiders. Random under sampling is used for reducing the class imbalance issue. The proposed approach is evaluated using the MobileNetV2, VGG19, and ResNet50 pre-trained models, and a benchmark dataset. Experimental results show that the proposed method is effective and outperforms other state-of-the-art methods.

show abstract

Section: Proposed Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Image-Based Feature Representation for Insider Threat Classification

2020

View full text Add to dashboard Cite

show abstract

“…The proposed system adopts applying transfer learning [20,21] using the well-known CNN models, i.e., AlexNet (Figure 4, [22]) and ResNet50 ( Figure 5, [23]). Transfer learning is based on using pre-trained models that have been trained over a huge training database.…”

Section: Deep-learning Architecturementioning

confidence: 99%

“…On the other hand, AlexNet is composed of five convolutional layers and three fully connected layers (see Figure 4). More details about AlexNet and ResNet models can be found in [22] and [23], respectively.…”

Section: Deep-learning Architecturementioning

confidence: 99%

See 1 more Smart Citation

Deep Learning Grading System for Diabetic Retinopathy using Fundus Images. (Dept. E)

Elsawah¹,

Elnakib²,

Moustafa³

2020

MEJ. Mansoura Engineering Journal

View full text Add to dashboard Cite

Diabetic Retinopathy (DR) is one of the main causes of blindness that can be overcome, if it is early detected. This work proposes an automated early detection and grading of DR using fundus images. The proposed detection and grading system investigate different deep learning architectures (i.e., ResNet and AlexNet) that are applied to an augment data to extract deep compact features of the fundus images. The extracted features are input to a pixel-wise Neural Network (NN) classifier or a Support Vector Machine (SVM) classifier for automated DR grading. The performance of the proposed system is evaluated using a publically available fundus Indian Diabetic Retinopathy Image Dataset (IDRiD), collected for ISBI-2018 challenge. The IDRiD dataset consists of 516 retinal images of normal and different DR grades, i.e., mild, moderate, severe, and Proliferative Diabetic Retinopathy (PDR). Our system achieves an overall accuracy of 95.73%, sensitivity of 95.73%, and specificity of 98.51% utilizing an AlexNet-based architecture and a pixel-wise NN classifier. Compared to the previous related work, the proposed system shows promising DR grading performance.

show abstract