Diseases of grape leaves are destructive because they spread very quickly. Therefore, they should be detected and tackled fast. This study proposes a novel Enhanced VGG16 model for detecting, identifying, and classifying three widespread diseases in grape leaves. These diseases include Downey Mildew, Anthracnose, and Powdery Mildew. The data are automatically gathered with the assistance of photos taken with a quadcopter from the grapevine garden. The main central system analyzes the images received from the quadcopter to identify diseases. If it finds any of the three diseases, it sends the command to the hexacopter to spray pesticide into the grapevine garden location where the disease is found. The proposed method using the Enhanced VGG16 model with Faster Region-based Convolutional Neural Networks (R-CNN) is compared with different networks, including VGG16, GoogLeNet, ResNet50, and AlexNet. The experimental results on the grape leaf diseases demonstrated that the proposed method achieves the mean Average Precision) mAP (criterion improvements of 0.53%, 0.912%, 2.759%, and 7.268% compared with the ResNet50, VGG16, GoogLeNet, and AlexNet networks, respectively. Also, the average accuracy of the proposed Enhanced VGG16 model is 99.6%, which is 0.437–1.91% higher than other models. The Enhanced VGG16 has the best precision, and the number of layers is acceptable, although it is not less.
19 August 2021 in Security and Communication Networks
The characteristics of the mobile ad hoc network (MANET), such as no need for infrastructure, high speed in setting up the network, and no need for centralized management, have led to the increased popularity and application of this network in various fields. Security is one of the essential aspects of MANETs. Intrusion detection systems (IDSs) are one of the solutions used to ensure security in this network. Clustering-based IDSs are very popular in this network due to their features, such as proper scalability. This paper proposes a new algorithm in MANETs to detect black hole attack using the K-nearest neighbor (KNN) algorithm for clustering and fuzzy inference for selecting the cluster head. With the use of beta distribution and Josang mental logic, the trust of each node will be calculated. According to the reputation and remaining energy, fuzzy inference will select the cluster head. Finally, the trust server checks the destination node. If allowed, it notifies the cluster head; otherwise, it detects the node as a malicious node in the black hole attack in each cluster. The simulation results show that the proposed method has improved the packet loss rate, throughput, packet delivery ratio, total network delay, and normalized routing load parameters compared with recent black hole detection methods.
22 September 2020 in Security and Communication Networks
One of the important issues in the computer networks is security. Therefore, trusted communication of information in computer networks is a critical point. To have a safe communication, it is necessary that, in addition to the prevention mechanisms, intrusion detection systems (IDSs) are used. There are various approaches to utilize intrusion detection, but any of these systems is not complete. In this paper, a new cross-correlation-based feature selection (CCFS) method is proposed and compared with the cuttlefish algorithm (CFA) and mutual information-based feature selection (MIFS) features with use of four different classifiers: support vector machine (SVM), naive Bayes (NB), decision tree (DT), and K-nearest neighbor (KNN). The experimental results on the KDD Cup 99, NSL-KDD, AWID, and CIC-IDS2017 datasets show that the proposed method has a better performance in accuracy, precision, recall, and F1-score criteria in comparison with the other two methods in different classifiers. Also, the results on different classifiers show that the usage of the DT classifier for the proposed method is the best.
Autocorrelation domain is a proper domain for clean speech signal and noise separation. In this paper, a method is proposed to decrease effects of noise on the clean speech signal, autocorrelation-based noise subtraction (ANS). Then to deal with the error introduced by assumption that noise and clean speech signal are uncorrelated, two methods are proposed. Also to improve recognition rate of speech recognition system, overestimation parameter is used. Finally, with the addition of energy and cepstral mean and variance normalization to features of speech, recognition rate has improved considerably in comparison to standard features and other correlation-based methods. The proposed methods are tested on the Aurora 2 database. Between different proposed methods, autocorrelation-based noise subtraction method with smoothing, overestimation, energy, and cepstral mean and variance normalization (ANSSOEMV) method has a best recognition rate improvement in average than MFCC features which is 64.91% on the Aurora 2 database.