Journal of Soft Computing and Computer Applications

Cyberattacks have become one of the most significant security threats that have emerged in the last couple of years. It is imperative to comprehend such attacks; thus, analyzing various kinds of cyberattack datasets assists in constructing the precise intrusion detection models. This paper tries
to analyze many of the available cyberattack datasets and compare them with many of the fields that are used to detect and predict cyberattack, like the Internet of Things (IoT) trac-based, network trac-based, cyber-physical system, and web track-based. In the present paper, an overview of each of them is provided, as well as the course of machine learning that has employed these datasets. From this survey, the researchers and the cybersecurity professional can derive a convenient classification of these datasets and their usages based on reviewing recent papers in this field. Furthermore, the types of machine learning involved in such systems as well as the intrusion detection and anomaly detection systems used to learn the models are presented in this paper. These techniques include deep learning models, random forests, support vector machines, and other commonly applied methods. Each technique has its advantages and limitations in the context of cyberattack prediction and detection. The paper is also consider factors like the specific technique and tools used, the type of attacks taken and the accuracy rate achieved. Of a total of 85 papers, 34 were selected for review in this paper. This survey is an essential tool for improving knowledge about the state of cyber detection and prediction techniques today.

Artificial neural networks play a crucial role in machine learning and there is a need to improve their performance. This paper presents FOXANN, a novel classification model that combines the recently developed Fox optimizer with ANN to solve ML problems. Fox optimizer replaces the backpropagation algorithm in ANN; optimizes synaptic weights; and achieves high classification accuracy with a minimum loss, improved model generalization, and interpretability. The performance of FOXANN is evaluated on three standard datasets: Iris Flower, Breast Cancer Wisconsin, and Wine. The results presented in this paper are derived from 100 epochs using 10-fold cross-validation, ensuring that all dataset samples are involved in both the training and validation stages. Moreover, the results show that FOXANN outperforms traditional ANN and logistic regression methods as well as other models proposed in the literature such as ABC-ANN, ABC-MNN, CROANN, and PSO-DNN, achieving a higher accuracy of 0.9969 and a lower validation loss of 0.0028. These results demonstrate that FOXANN is more effective than traditional methods and other proposed models across standard datasets. Thus, FOXANN effectively addresses the challenges in ML algorithms and improves classification performance.

Digital data such as images, audio, and video have become widely available since the invention of the Internet. Due to the ease of access to this multimedia, challenges such as content authentication, security, copyright protection, and ownership determination arose. In this paper,
an explanation of watermark techniques, embedding, and extraction methods are provided. It further discusses the utilization of artificial intelligence methods and conversion of host media from the spatial domain to the frequency domain; these methods aim to improve the quality of watermarks. This paper also included a classification of the basic characteristics of the digital watermark and the most recent areas in which the watermark was used. A comparative analysis of watermarking technologies has been conducted over the past few years (from 2018 to date), and a set of challenges are discussed, such as domain, robustness, security, and visual imperceptibility, as well as methods to address them. The findings of this paper demonstrate that employing intelligent methods can enhance the robustness of the watermark by making it undetectable and challenging to manipulate the host media.

In lightweight cryptography, the absence of an S-Box in some algorithms like speck, Tiny Encryption Algorithm, or the presence of a fixed S-Box in others like Advanced Encryption Standard can make them more vulnerable to attacks. This study introduces an innovative method for creating a dynamic 6-bit S-Box (8 x 8) in octal format. The generating process of S-Box passes through two phases. The first is the number initialization phase. This phase involves generating sequence numbers 1, sequence numbers 2, and sequence numbers 3 depending on Xi, Yi, and Zi
values generated using the 3D Hindmarsh Rose model. The second is the S-Box construction phase. This phase involves building S-Box values depending on sequence numbers 1, sequence numbers 2, and sequence numbers 3, which resulted from the number initialization phase. The e
effectiveness of the proposed S-Box was evaluated through various criteria, including the bijective property, balanced, fixed points, opposite fixed points, completeness criteria, avalanche criteria, and strict avalanche criteria. It was observed that S-Box achieved a linear and differential branch number of 3, non-linearity of 24, differential uniformity of 4, and algebraic degree of 3. In addition, reducing the number of linear and nonlinear operations makes it suitable for lightweight algorithms. The architecture of the proposed S-Box demonstrates robustness, with a total of 1.981087 possible S-Boxes against algebraic attacks. Moreover, the construction of the S-Box and its inverse take only 14.3542 milliseconds, making it suitable for use in many lightweight block ciphers. The new S-Box is the first to exhibit these characteristics.

The primary elements of Intelligent Transportation Systems (ITSs) have become Vehicular Ad-hoc NETworks (VANETs), allowing communication between the infrastructure environment and vehicles. The large amount of data gathered by connected vehicles has simplified how Deep Learning (DL) techniques are applied in VANETs. DL is a subfield of artificial intelligence that provides improved learning algorithms able to analyzing and process complex and heterogeneous data. This study explains the power of DL in VANETs, considering applications like decision-making,
vehicle localization, anomaly detection, trac prediction and intelligent routing, various types of DL, including Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs) are mentioned for their efficiency in VANET applications. The DL algorithms in VANETs have garnered attention from academia and industry, leading to the development of architectures and algorithms tailored for VANETs. The challenges and advantages of DL in VANETs are expected as future research directions in this field. Moreover, this study explains the operations of Swarm Intelligence (SI) techniques, such as Ant Colony Optimization (ACO), Stochastic Diffusion Search (SDS), Particle Swarm Optimization (PSO), and Artificial Swarm Intelligence (ASI) in VANETs.
The techniques of SI offer solutions for improving problems and can be utilized to diagnose and manage routing protocols and trac congestion malicious nodes in VANETs. This study offers a detailed diagnosis of how SI and DL help improve the efficiency and performance of VANETs. This improvement facilitates the development more safer and active transportation systems with intelligent capabilities.

Video anomaly detection is one of the trickiest issues in intelligent video surveillance because of the complexity of real data and the hazy definition of anomalies. Since abnormal occurrences typically seem different from normal events and move differently. The global optical flow was determined with the maximum accuracy and speed using the Farneback approach for calculating the magnitudes. Two approaches have been used in this study to detect strangeness in the video. These approaches are Deep Learning (DL) and manuality. The first method uses the activity map’s development of entropy to detect the oddity in the video using a particular threshold. The second method uses a Convolutional Recurrent Auto Encoder (CRAE). CRAE is a network that combines a convolutional autoencoder and an attention-based Convolutional Long-Short-Term Memory (ConvLSTM) network. The irregularity regarding the temporal pattern and the spatial irregularity, respectively, might be captured by the convolutional autoencoder and ConvLSTM network. The current output properties of each CovnLSTM layer were extracted from their hidden states using the attention method. Comparing the error with an experimentally established threshold, anomalies were specified to exist and a convolutional decoder was used to recreate the input video clip and the testing video clip. The best detection of whether in-frame variation was abnormal or normal, a trial-and-error threshold was 0.04 for handcrafted features through the University of Minnesota (UMN) dataset and 0.00035 for DL features through the avanue dataset.

The coronavirus disease 2019 outbreak caused widespread disruption. The World Health Organization has recommended wearing face masks, along with other public health measures, such as social distancing, following medical guidelines, and thermal scanning, to reduce transmission, reduce the burden on healthcare systems, and protect population groups. However, wearing a mask, which acts as a barrier or shield to reduce transmission of infection from infected individuals, hides most facial features, such as the nose, mouth, and chin, on which face detection systems depend, which leads to the weakness of these systems. This paper aims to provide essential insights for researchers and practitioners interested in developing and implementing deep learning-based face mask detection systems. Although current deep learning models have made significant strides and tremendous advances in many applications, including security, access control, and identity verification, development efforts continue. This paper also discusses the importance of the datasets used to train and evaluate these models, emphasising the need for diverse data such as mask types, facial occlusions, lighting conditions, and high-quality data to enhance model performance. In addition to the challenges posed by real-world conditions that can a
ect detection accuracy, face detection and face mask recognition methods were compared to deep learning models, where the accuracy of the Multi-Task ArcFace model reached 99.78, which is the highest accuracy among other detection methods.

The Bidirectional Long Short-Term Memory (Bi-LSTM) network structure enables data analysis, enhances decision-making processes, and optimizes resource allocation in cloud computing systems. However, achieving peak network performance relies heavily on choosing the hyperparameters for configuring the network. Enhancing resource allocation improves the Service Level Agreement (SLA) by ensuring efficient utilization and allocation of computational resources based
on dynamic workload demands. This paper proposes an approach that integrates a Multi-Objective Evolutionary Algorithm (MOEA) with deep learning techniques to address this challenge. This approach combines the optimization capabilities of MOEA with the learning predictive models to establish a framework for resource allocation in cloud environments. Snake Optimizer Algorithm (SOA) and Genetic Algorithm (GA) are utilized to identify specific hyperparameters through optimization procedures, which serve as the foundation for setting up the Long Short-Term Memory (LSTM) network. The search process begins with a random selection of parameters. The process is iterative in nature and concentrates on the best results for the parameters, with initial random sampling done according to the results obtained. This methodology considers all areas and examines the most effective areas, which improves the eeffctiveness of the approach when solving the problem of moving through the hyperparameter space. The modifications that were made to the snake algorithm have the goal of optimizing resource management within the data center in cloud environment. LSTM-SOA approach achieved an accuracy level of 97% on the dataset created in the Cloudsim simulation environment. This approach is based on multi-objectives to meet the requirements of determining the optimal data center. By comparison, the LSTM-AG approach achieved an accuracy level of 87%.

Multi-Objective Optimization (MOO) techniques have become increasingly popular in recent years due to their potential for solving real-world problems in various fields, such as logistics, finance, environmental management, and engineering. These techniques offer comprehensive
solutions that traditional single-objective approaches fail to provide. Due to the many innovative algorithms, it has been challenging for researchers to choose the optimal algorithms for solving their problems. This paper examines recently developed MOO-based algorithms. MOO is introduced along with Pareto optimality and trade-o analysis. In real-world case studies, MOO algorithms address complicated decision-making challenges. This paper examines algorithmic methods, applications, trends, and issues in multi-objective optimization research. This exhaustive review explains MOO algorithms, their methods, and their applications to real-world problems. This paper aims to contribute further advancements in MOO research. No singular strategy is superior; instead, selecting a particular method depends on the natural optimization problem, the computational resources available, and the specific objectives of the optimization tasks.