Search

The SARS-CoV-2 virus causes a respiratory disease in humans, known as COVID-19. The confirmatory diagnostic of this disease occurs through the real-time reverse transcription and polymerase chain reaction test (RT-qPCR). However, the period of obtaining the results limits the application of the mass test. Thus, chest X-ray computed tomography (CT) images are analyzed to help diagnose the disease. However, during an outbreak of a disease that causes respiratory problems, radiologists may be overwhelmed with analyzing medical images. In the literature, some studies used feature extraction techniques based on CNNs, with classification models to identify COVID-19 and non-COVID-19. This work compare the performance of applying pre-trained CNNs in conjunction with classification methods based on machine learning algorithms. The main objective is to analyze the impact of the features extracted by CNNs, in the construction of models to classify COVID-19 and non-COVID-19. A SARS-CoV-2 CT data-set is used in experimental tests. The CNNs implemented are visual geometry group (VGG-16 and VGG-19), inception V3 (IV3), and EfficientNet-B0 (EB0). The classification methods were k-nearest neighbor (KNN), support vector machine (SVM), and explainable deep neural networks (xDNN). In the experiments, the best results were obtained by the EfficientNet model used to extract data and the SVM with an RBF kernel. This approach achieved an average performance of 0.9856 in the precision macro, 0.9853 in the sensitivity macro, 0.9853 in the specificity macro, and 0.9853 in the F1 score macro.

Nowadays, the high number of devices and applications connected to the Internet has generated a great amount of data being which makes privacy and protection a more challenging task. In addition, new technologies, such as the Internet of Things, incorporate many resource-constrained devices in the network. Reliable cryptography algorithms have to be employed to deal with this problem, which also needs to be efficiently implemented in small devices. There are several algorithms for this purpose, among them, neural cryptography. In this context, this work proposes the implementation of an artificial neural network architecture called tree parity machine (TPM) to perform the exchange of keys through the mutual learning of these networks. This method is not based on number theory, which makes it less computationally costly, and can be an alternative for embedded systems, which generally have several limitations in the processing capacity and resources used. In the area of embedded systems, FPGAs have gained more space, thanks to their reconfiguration capacity. Thus, different methods for implementing a TPM in FPGA were tested and analyzed, in order to optimize the following performance parameters, the response time, the maximum frequency of operation, and the consumed area of the FPGA considering logical elements, embedded multipliers, and registers. In addition, software implementation based on a multi-core CPU was used for comparison purposes. Experimental results demonstrated that the implementation of parallelism in FPGA for different blocks of the TPM weight matrix reached the best performance results. Thus, our proposal intends to develop an economic component in terms of resource consumption, however, maintaining the characteristic of high processing capacity. Therefore, the methodologies presented in this paper intends to be a useful reference to optimize future implementations in FPGA for cryptography applications.

Currently, with the increasing number of devices connected to the Internet, search for network vulnerabilities to attackers has increased, and protection systems have become indispensable. There are prevalent security attacks, such as the Distributed Denial of Service (DDoS), which have been causing significant damage to companies. However, through security attacks, it is possible to extract characteristics that identify the type of attack. Thus, it is essential to have fast and effective security identification models. In this work, a novel Intrusion Detection System (IDS) based on the Tree-CNN hierarchical algorithm with the Soft-Root-Sign (SRS) activation function is proposed. The model reduces the training time of the generated model for detecting DDoS, Infiltration, Brute Force, and Web attacks. For performance assessment, the model is implemented in a medium-sized company, analyzing the level of complexity of the proposed solution. Experimental results demonstrate that the proposed hierarchical model achieves a significant reduction in execution time, around 36%, and an average detection accuracy of 0.98 considering all the analyzed attacks. Therefore, the results of performance evaluation show that the proposed classifier based on Tree-CNN is of low complexity and requires less processing time and computational resources, outperforming other current IDS based on machine learning algorithms.