Fake news has become a major issue owing to its quick growth on the internet, difficulty in identifying it from true news, and people's reliance on social media platforms as primary sources. It has negative effects at several levels, including individual, communal, political, and economical. Detecting Arabic fake news requires significant effort owing to limited datasets and studies in the sector. In this study, Employ Perturbation Adversarial attacks was applied as a regularization technique for fake news classification. Adversarial examples are generated by perturbing the model's word embedding matrix. The AraBERTv2 model is utilized for preprocessing operations. Five different machine learning models were trained on clean data and tested using both clean and adversarial examples to evaluate the generalization capabilities of the classification models. To address the scarcity of Arabic datasets, A translated English fake news was utilized dataset Experimental results indicate that the LightGBM and AdaBoost algorithms exhibited the best performance (90%) compared to other classifiers.

the rapid spread of fake news online, coupled with the difficulty of distinguishing it from real news, has become a serious issue, especially with social media being a primary news source for many people. The fake news can have damaging effects on individuals, communities, and political and economic systems. Detecting Arabic fake news presents additional challenges due to the limited availability of relevant datasets and research. In this study, Perturbation Adversarial attacks was applied as a regularization technique for fake news detection, generating adversarial examples by modifying the model’s word embedding matrix. The AraBERTv2 model is used for preprocessing, and the lack of Arabic data was overcome by utilizing a translated English fake news dataset. A Recurrent Neural Network (RNN) model was trained on clean data, testing it with both clean and adversarial examples to evaluate its generalization capability. The results demonstrate that the RNN model performs effectively, achieving strong accuracy in Arabic fake news detection.

Android is a widely used mobile platform with over 2 billion active users worldwide, which makes it is an ideal target for harmful actions such as malicious software programs production and distribution. Malware creators and developers are progressively using complicated and evasive anomalies. As a result, typical malware detection systems have become vulnerable to these attacks. As the threat landscape grows, the application of Artificial Intelligence (AI) technologies in anomaly detection has attracted the attention of a growing number of professionals and academics. The influence of AI techniques on anomaly detection in the Android environment is investigated in this study. Machine learning, deep learning and data mining techniques are being examined into this work as an AI-based methodologies for anomaly detection on Android. We investigate the advantages, disadvantages and challenges associated with various methodologies. We present also future research opportunities in this paper. The survey aims to offer a full understanding of the impact of AI on anomaly detection on Android.

Face recognition technology has been significantly impacted by the COVID-19 epidemic, among other parts of daily life. Facial recognition technologies, which were formerly employed for security, access control, and identity verification purposes, have actually become less reliable as a result of mask use. Although deep learning-based methods have demonstrated considerable promise in this field, it is still unknown how well various architectural approaches perform. The performance of five well-known deep learning architectures—ResNet50, VGG16, InceptionV3, MobileNetV2, and Xception—is compared in this study. We evaluate the architectures using a sizable dataset (LFW) made up of masked faces of various racial and gender compositions. According to our findings, the ResNet50 has a higher accuracy score of 98.3% compared to the Vgg16 architecture’s 98.2%, Xception’s 98, MobileNet’s 97.6 and Inception’s 97.4%. In order to determine the impact of various model elements, such as the quantity of convolutional layers, the quantity of parameters, and the time inference, we also perform an analysis study. Our findings show that a key factor in getting high accuracy for masked face recognition is the network’s depth.

Social media platforms generate vast amounts of data that can be mined to gain insights into user behavior, preferences, and opinions. With the widespread adoption of mobile devices and the Android platform, social media data mining on Android has become an important area of research. In this paper, we present a systematic review of existing research related to social media data mining on Android. We conducted a comprehensive literature search and identified relevant research papers published between 2015 and 2022. We analyzed the papers to identify key themes and trends, and we discuss the strengths and weaknesses of current approaches. Our review revealed that most of the research in this area focuses on text data mining, and that machine learning techniques are commonly used for data analysis. We also found that there is a lack of research on image and video data mining, and that privacy and ethical issues are important considerations in this field. Based on our review, we propose future research directions for social media data mining on Android, including the development of more efficient and accurate machine learning algorithms, and the integration of multimedia data analysis. Our paper provides a valuable overview of existing research in this field and offers insights for researchers and practitioners interested in social media data mining on Android.

This paper presents a new text independent speaker recognition system based on new cepstral Coefficients and i-vector based on probabilistic linear discriminant analysis. This system is designed to be used as a biometric authentication system based on the speaker voice. The experiments were performed on speech data taken from a large number of users and consist of 630 speakers from TIMIT database for different training conditions. The efficiency of the proposed system is observed with the increase of Speaker identification Rate. Experimental results show that the system gives higher recognition performance for different numbers of speakers.

Handwritten recognition has received considerable attention in the domain of pattern recognition, image processing, over the last few decades. As a consequence of this research effort, several algorithms were developed using different techniques. Particularly, Deep Learning has shown a remarkable capability to handle handwritten recognition in very recent years. The well-known Deep learning techniques are the Convolutional Neuronal Networks (CNNs) and Recurrent Neuronal Networks (RNNs). This paper provides a survey of the most recent handwritten recognition systems. Thus, we present the most significant algorithms for handwritten character/word/text recognition by explaining the different approaches used in the recognition process and we compare them in terms of accuracy.

An important amount of speech is typically required for speaker identification system development and evaluation. Nowadays, robust speaker identification systems when short utterances are used remains a key consideration for automatic speaker recognition, since a lot of real world applications are able to deal with only limited duration speech data. This paper presents a new approach based on a low complexity solution based on a new feature vectors to build Gaussian Mixture Models (GMM) for speaker identification systems especially when training and testing utterance lengths are reduced. We compared our proposed system to the state-of-the-art based system in Speaker identification. Experiments on TIMIT database were conducted to demonstrate that this new feature vector can outperform the standard GMM-based system and show that there is no need for extra-data to identify the speakers.

Recent advances in the speaker recognition (SR) field showed remarkably accurate and outperforming algorithms. However, their performances drastically degrade when the sparse amount of data is available. Nowadays, recognising a speaker identity when only a small amount of speech data is involved for testing and training remains a key consideration since many real world applications often have access to only speech data having a limited duration. In this study, the authors present a new improved approach, based on new information detected from the speech signal, to improve the task of automatic speaker identification. In doing so, they highlight how the detection of the speaker dialect can be explored to address the research problem related to short utterance SR. Results obtained with the new regional system are presented which provide a comparison between this system and the state‐of‐the‐art systems for speaker identification task.

Thresholding is a popular image segmentation method that converts gray-level image into binary image. The problem of thresholding has been quite extensively studied for many years in order to get an optimum threshold value. The multi-level thresholding becomes very computationally challenges. In this paper, a novel multilevel thresholding method based on particle swarm optimization (PSO) algorithm is proposed, or it seems to be the best tool, to maximize the Kapur and Otsu objective functions. We employed the properties of discriminate analysis using Kapur and Otsu methods to render the optimal thresholding technics more applicable and effective. The obtained result and the comparative study illustrate the algorithm's outstanding performances in segmenting both the grey level image and the MRI scans.

Over the past decade, the field of automatic speaker recognition has been the subject of extensive research looking for an efficient determination of a person's identity. Despite the essential role played by acoustic characteristics in order to discriminate between speakers. The research of discriminative information about a person remains a major challenge. The main objective of this paper is to present a new approach employing additional information which is dialect detection with a novel parameterization of the speech to improve the task of speaker identification. The superiority of the proposed system has been demonstrated by different kernels function of Support Vector Machines (SVM) with speakers taken from TIMIT database.

Gaussian mixture models (GMM) have become the standard method used for speaker recognition systems. A recent discovery is that combining GMM approach with another classifier is an effective method for speaker classification. We consider the GMM supervector in the context of support vector machines (SVM). We construct a support vector machine tested with two kernel functions employing the GMM supervectors. The main idea of the study is to combine the discriminative classifier SVM and the traditional GMM pattern classification with a new dimensional cepstral feature vector extracted from the speech to achieve better classification rate. This idea has been analytically formulated and tested on speakers from TIMIT database. First we describe the SVM-GMM system then we briefly discuss how the new low dimensional feature vector can feed to identification rate. We show comparative results obtained with GMM, SVM, GMM-SVM based system and existing works. Thereafter, we show that the new hybrid system can outperform the standard GMM-SVM based system and give remarkable increases in speaker identification rates.