Abstract—Over the last few decades, rapid progress in AI, machine learning, and deep learning has resulted in new techniques and various tools for manipulating multimedia. Though the technology has been mostly used in legitimate applications such as for entertainment and education, etc., malicious users have also exploited them for unlawful or nefarious purposes. For example, high-quality and realistic fake videos, images, or audios have been created to spread misinformation and propaganda, foment political discord and hate, or even harass and blackmail people. The manipulated, high-quality and realistic videos have become known recently as Deepfake. Various approaches have since been described in the literature to deal with the problems raised by Deepfake. To provide an updated overview of the research works in Deepfake detection, we conduct a systematic literature review (SLR) in this paper, summarizing 112 relevant articles from 2018 to 2020 that presented a variety of methodologies. We analyze them by grouping them into four different categories: deep learning-based techniques, classical machine learning-based methods, statistical techniques, and blockchain-based techniques. We also evaluate the performance of the detection capability of the various methods with respect to different datasets and conclude that the deep learning-based methods outperform other methods in Deepfake detection.

Abstract—Android is the most well-known portable working framework having billions of dynamic clients worldwide that pulled in promoters, programmers, and cybercriminals to create malware for different purposes. As of late, wide-running inquiries have been led on malware examination and identification for Android gadgets while Android has likewise actualized different security controls to manage the malware issues, including a User ID (UID) for every application, and framework authorization. In this paper, we advance and assess various kinds of machine learning (ML) by applying ensemble-based learning systems for identifying Android malware related to a substring-based feature selection (SBFS) strategy for the classifiers. In the investigation, we have broadened our previous work where it has been seen that the ensemble-based learning techniques acquire preferred outcomes over the recently revealed outcome by directing the DREBIN dataset, and in this manner gives a solid premise to building compelling instruments for Android malware detection.

Abstract—In recent years, a widespread research is conducted with the growth of malware resulted in the domain of malware analysis and detection in Android devices. Android, a mobile-based operating system currently having more than one billion active users with a high market impact that have inspired the expansion of malware by cyber criminals. Android implements a different architecture and security controls to solve the problems caused by malware, such as unique user ID (UID) for each application, system permissions, and its distribution platform Google Play. There are numerous ways to violate that fortification, and how the complexity of creating a new solution is enlarged while cybercriminals progress their skills to develop malware. A community including developer and researcher has been evolving substitutes aimed at refining the level of safety where numerous machine learning algorithms already been proposed or applied to classify or cluster malware including analysis techniques, frameworks, sandboxes, and systems security. One of the most promising techniques is the implementation of artificial intelligence solutions for malware analysis. In this paper, we evaluate numerous supervised machine learning algorithms by implementing a static analysis framework to make predictions for detecting malware on Android 

Abstract—The distributed database system is the combination of two fully divergent approaches to data processing: database systems and computer network to deliver transparency of distributed and replicated data. The key determination of this paper is to achieve data integration and data distribution transparency, study and recognize the problems and approaches of the distributed database system. The distributed database is evolving technology to store and retrieve data from several location or sites with maintaining the dependability and obtain ability of the data. In the paper we learn numerous problems in distributed database concurrency switch, design, transaction management problem etc. Distributed database allows to end worker to store and retrieve data anywhere in the network where database is located, during storing and accessing any data from distributed database through computer network faces numerous difficulties happens e.g. deadlock, concurrency and data allocation using fragmentation, clustering with multiple or single nodes, replication to overcome these difficulties it is essential to design the distributed database sensibly way. 

Abstract—In the digital world, inpainting is the algorithm used to replace or reconstruct lost, corrupted, or deteriorated parts of image data. Of the various proposed inpainting methods, convolution methods are the simplest and most efficient. In this paper, an enhanced inpainting model based on convolution theorem is proposed for digital images that preserves the edge and effectively estimates the lost or damaged parts of an image. In the proposed algorithm, a mask image is created dynamically to detect the image area to inpaint where most of the algorithms detect the missing parts of the image manually. Studies confirm the simplicity and effectiveness of our method, which also produces results that are comparable to those produced using other methods. 

Abstract—This is a preliminary study and the objective of this study has been to compare the performance of some of the primitive and fundamentally different post acquisition image enhancement algorithms as applied to different ultrasound (US) images. Such a comparison would help to decide as to which algorithm could be useful for clinicians, and in evaluating the role of different US images enhancement in a soft-copy environment. In this study, 3 types of US images (Liver, kidney & Abdomen) were taken, and 5 fundamentally different and widely employed image enhancement techniques were applied on these images. As the principal objective of image enhancement is to obtain an image with a high content of visual detail, a multi point rank-order method was used to identify small differences or trends in observations. Among the different algorithms, the proposed modified Anisotropic diffusion filtering outperformed than other techniques . 

Abstract—Android, a most popular mobile operating system having more than billions of active users with a high market impression that have encouraged the cyber-criminals to push the malware into this operating system. In recent years, an extensive research is conducted in the domain of malware analysis and detection in Android devices. And Android already have developed and implemented numerous security controls to solve the problems. In this paper, we apply a new Blockchain technology to evaluate and exchange various machine learning model for a particular dataset by interacting with smart contracts that offer a reward. This allows contributors to submit their solution to the Blockchain by training with selected machine learning models for a reward in a trustless manner. This experimentation leads to provide a strong basis for building effective tools for Android malware detection. 

Abstract—Android is a most popular mobile-based operating system with billions of active users, which has encouraged hackers and cyber-criminals to push the malware into this operating system. Accordingly, extensive research has been conducted on malware analysis and detection for Android in recent years; and Android has developed and implemented numerous security controls to deal with the problems, including unique ID (UID) for each application, system permissions, and its distribution platform Google Play. In this paper, we evaluate four tree-based machine learning algorithms for detecting Android malware in conjunction with a substring-based feature selection method for the classifiers. In the experiments 11,120 apps of the DREBIN dataset were used where 5,560 contain malware samples and the rest are benign. It is found that the Random Forest classifier outperforms the best previously reported result (around 94% accuracy, obtained by SVM) with 97.24% accuracy, and thus provides a strong basis for building effective tools for Android malware detection. 

Abstract—Numerous technological improvements have found in the academic setting which removes the binding of educators and students from time and space. Day by day, the rate of drop-out is increasing by separation of them from learning. The main effort of our project is to fulfill a mission allowing individuals to learn or educate without physically attending. In this paper, we build an innovative web-based application enabling the teachers and students with numerous educational exercises using computer/smart devices. Using this application teacher, students and parents can collaborate on a single podium, while teachers can counsel with students in a real-time and share the performance and actions with parents as well as administrators. A method to modification of our traditional education system but not the replacement of teaching, it’s only the enhancement idea for teaching helps to learn easily and fill up their liking. 

Abstract—In the rapidly evolving landscape of artificial intelligence (AI), particularly in the Deepfake domain, largescale datasets play a pivotal role in ensuring performance, including the model’s accuracy, robustness, trustworthiness, etc. However, the increasing size and intricacy of the datasets impose a growing demand for computational resources and amplify the cost and duration of model building. To mitigate the challenge, dataset distillation provides a solution. For the Deepfake detection problem, noteworthy datasets such as VDFD, FaceForensics++, DFDC, and Celeb-DF underscore the indispensability of extensive data for ensuring model robustness. Nevertheless, the computational requirement associated with these datasets presents significant obstacles. This paper describes a data distillation method utilizing Active Learning to reduce dataset size while retaining essential data qualities. The proposed method facilitates efficient model training selecting representative samples by capturing the most salient features, thereby enabling effective performance in resource-constrained environments. The study encompasses developing a data distillation algorithm tailored for Deepfake detection, rigorous experimentation with a major Deepfake dataset to validate its efficacy, and a comprehensive comparison of the model performance trained on distilled versus original datasets. Through thorough analysis, we demonstrate the practicality and effectiveness of our proposed method in alleviating computational demands without compromising detection accuracy.

Abstract—As techniques and tools for synthetic media and Deepfakes continue to advance, it is increasingly clear that video, audio, and images can no longer be relied upon as truthful recordings of reality. Every digital communication channel is now vulnerable to manipulation, and there is widespread use of Deepfakes to propagate misinformation and disinformation, inflame political discord, defame opposition, commit cyber frauds, or blackmail individuals. While deep learning (DL) methods have been widely used to identify Deepfakes, this paper demonstrates that classical machine learning (ML) methods can achieve superior performance--comparable with or exceeding state-of-the-art DL methods in detecting Deepfakes. Using the traditional procedures of feature development and selection, training, and testing of ML classifiers for the task actually provides better understandability and interpretability while consuming much less computing resources. In addition, an omnibus test, the Analysis of Variance (ANOVA), is conducted to compare the performance of multiple ML models. We present experiments that achieve 99.84% accuracy on the FaceForecics++ dataset, 99.38% accuracy on the DFDC dataset, 99.66% accuracy on the VDFD dataset, and 99.43% accuracy on the Celeb-DF dataset. Our study thus challenges the notion that DL approaches are the only effective way to detect Deepfakes and demonstrates that judicious use of ML approaches can be highly efficacious and cost-effective.

Abstract—In recent years, the rapid advancements in machine learning (ML), artificial intelligence (AI), and deep learning (DL) have ushered in a new era of sophistication in image and video manipulation techniques. Notably, the emergence of Deepfake technology, driven by AI, has garnered substantial attention. Deepfakes involve training DL models on extensive datasets of similar faces and subsequently mapping one person's expressions onto another's face, resulting in deceptively realistic fake videos that can be virtually indistinguishable from authentic ones. The proliferation of Deepfakes poses various threats, including the potential to incite political turmoil, coerce individuals, or fabricate false terrorist incidents. This trend undermines privacy, consent, and the trustworthiness of digital media. Consequently, there is a pressing need to continually advance Deepfake detection and prevention methodologies to safeguard against their malevolent use and uphold the integrity of digital content. This paper introduces a Convolutional Neural Network (CNN) based DL model specifically developed for the classification of Deepfake video frames. Our model exhibits impressive performance, as validated by a thorough analysis of the VDFD dataset, where it achieves an outstanding average precision, recall, and F1-score of 95%, 94%, and 94%, respectively. Moreover, our model showcases its efficacy across various widely recognized Deepfake datasets, including FF++, Celeb-DF, and DFDC, with frame detection average rates for precision, recall, and F1-score ranging from 80% to 85%. These compelling results signify that our proposed CNN-based frame detection technique is a powerful tool for Deepfake detection, emphasizing the critical significance of automated Deepfake detection with an exceptionally high detection rate. This technology represents a pivotal step toward protecting against the potential misuse of Deepfakes, reinforcing the security and integrity of digital content in our modern world.

Abstract—In recent years, advances in technology, specifically in machine learning, artificial intelligence, and deep learning, have made it possible to create highly convincing fake videos known as Deepfakes. These videos are produced by training computer models on extensive datasets of faces and then seamlessly blending one person’s facial expressions onto another’s, resulting in videos that are nearly indistinguishable from reality. The widespread use of Deepfakes presents several concerns, including the creation of political distress, the occurrence of fake terrorism events, the undermining of trust in digital media, etc. Therefore, there is an urgent need to continually advance Deepfake detection and prevention methodologies to safeguard against their malevolent use and maintain the integrity of digital content. This paper conducts a meticulous analysis of the current landscape of Deepfake research, surveys the most effective detection solutions, and introduces a real-time Deepfake detection and prevention model within a rigorous testing framework. This model integrates innovative Blockchain and Steganalysis technologies to provide a robust solution to combat the explosion of Deepfakes. Our holistic framework offers a systematic and statistically rigorous approach to distinguishing genuine content from its manipulated counterparts, Deepfakes. By employing the principles of hypothesis testing, and a robust test statistic, our research equips us with the analytical tools necessary to make well-informed and precise classifications, significantly contributing to the ongoing battle against Deepfakes.

Abstract—In the challenging domain of engineering, where cold regions present formidable challenges, we confront the relentless forces of nature. From sub-zero temperatures to the unpredictable dance of snowfall and the silent buildup of ice, these regions demand innovative solutions to fortify the resilience of critical infrastructure. This initiative harnesses the potential of cutting-edge technology and leverages the extensive historical weather data tapestry. It introduces a pioneering strategy by integrating machine learning algorithms with extensive weather data, steering cold region engineering into an era defined by foresight and adaptability. This paper studies a transformative approach designed to forecast, prevent, and ultimately enhance infrastructure resilience in the face of rigid cold. Addressing the distinct challenges of cold region engineering, arising from harsh winter conditions such as extreme cold temperature, snowfall, and ice accumulation, we offer a comprehensive study using machine learning algorithms applied to historical weather data to construct a deeper analysis model capable of highlighting adverse weather effects. This, in turn, covers the way for optimized resource allocation, streamlined maintenance planning, and design enhancements. Our proposed study follows a systematic process, encompassing meticulous data collection, appropriate feature selection, and aiming seamless integration of the model into existing infrastructure management systems. Additionally, it facilitates the implementation of efficient and proactive measures to mitigate the impact of severe weather conditions on infrastructure. The paper also conducts three different hypotheses testing: temperature impact hypothesis, precipitation influence hypothesis, and ice accumulation and infrastructure resilience hypothesis, propelling engineering practices to new heights, particularly in the face of challenging cold environments.

Abstract—This paper reviews security testing, covering the concepts and notions essential to the discipline. The following analysis uses an inverted funnel approach to reduce the broad topic of software testing into its constituent parts. With this framework in mind, the paper will attempt to answer the following questions. Why is software testing significant? What are the roles of users and developers as they relate to software? What techniques are commonly used when testing software? Is the discipline of software testing unified; if not, what controversies exist, and what do they concern? What is security testing? What is the purpose of security testing? What metrics are used to guide security testing? What are common threats that hamper or impede security testing? How can developers mitigate the risk posed by security threats; what defensive options are available? What improvements can be made to software testing? What can occur in the absence of software/security testing?

Abstract—Vehicular traffic and congestion are a major challenge worldwide because of the rapid growth in urban population. The congestion can be mitigated to enhance traffic management by predicting the accurate travel time of the vehicles in the traffic. This research developed a novel methodology utilizing machine learning on real-time traffic data collected through Bluetooth sensors deployed at traffic intersections to estimate travel time. The research evaluates the performance and accuracy of five different prediction systems for travel time estimation highlighting the effectiveness of the machine learning models in accurately predicting travel time. The research also explores the development of the machine learning model predicting the travel time during peak hours, considering traffic lights impact on travel time between intersections. The research findings contribute to the efficient and reliable travel time prediction systems development, helping commuters making informed decisions and improve traffic management strategies.

Abstract—This tutorial presents developments on the detection of Deepfakes, which are realistic images, audios and videos created using deep learning techniques. Deepfakes can be readily used for malicious purposes and pose a serious threat to privacy and security. The tutorial summarizes recent Deepfake detection techniques and evaluates their effectiveness with respect to several benchmark datasets. Our study finds that no single method can reliably detect all Deepfakes and, therefore, combining multiple methods is often necessary to achieve high detection rates. The study also suggests that more extensive and diverse datasets are needed to improve the accuracy of detection algorithms. A taxonomy of Deepfake detection techniques is introduced to aid future research and development in the field. We conclude by calling for the development of more effective Deepfake detection methods and countermeasures to combat this evolving and spreading threat. 

Abstract—Android operating system is one of the most prominent operating systems among mobile device users worldwide. But it is often the most targeted platform for malicious activities. Many researchers have studied android malware detection systems over the previous years. But android malware detection systems face many challenges and obfuscation is one of them. String encryption is one such obfuscation technique that helps android malware to evade malware detection systems. To address this challenge in android malware detection systems, a novel approach is being proposed in this study where Crypto-Detector: an open-source cryptography detection tool has been used in decompiled application code to extract encrypted strings and encryption methods as features. Accuracy of 0.9880 and f1-score of 0.9843 have been achieved during performance evaluation. The performance of our framework has been compared to those of other similar existing works and our work has outperformed all of them. 

Abstract—Deepfake, a new video manipulation technique, has drawn much attention recently. Among the unlawful or nefarious applications, Deepfake has been used for spreading misinformation, fomenting political discord, smearing opponents, or even blackmailing. As the technology becomes ever more sophisticated and the apps for creating them more readily available, detecting Deepfake has become a challenging task and researchers have proposed various deep learning (DL) methods for detection. Though the DL-based approach can achieve good solutions, this paper presents the results of our study indicating that traditional machine learning (ML) techniques alone can obtain superior performance in detecting Deepfake. The ML-based approach is based on the standard methods of feature development and feature selection, followed by training, tuning, and testing an ML classifier. The advantage of the ML approach is that it allows better understandability and interpretability of the model with reduced computational cost. We present results on several Deepfake datasets that are obtained relatively fast with comparable or superior performance to the state-of-the-art DLbased methods: 99.84% accuracy on FaceForecics++, 99.38% accuracy on DFDC, 99.66% accuracy on VDFD, and 99.43% on Celeb-DF datasets. Our results suggest that an effective system for detecting Deepfakes can be built using traditional ML methods. 

Abstract—Recent  advances  in  technology  have  made  the  deep  learning  (DL)  models  available  for  use  in  a  wide  variety  of  novel    applications;    for    example,    generative    adversarial    network   (GAN)   models   are   capable   of   producing   hyper-realistic  images,  speech,  and  even  videos,  such  as  the  so-called  “Deepfake” produced by GANs with manipulated audio and/or video  clips,  which  are  so  realistic  as  to  be  indistinguishable  from the real ones in human perception. Aside from innovative and  legitimate  applications,  there  are  numerous  nefarious  or  unlawful ways to use such counterfeit contents in propaganda, political  campaigns,  cybercrimes,  extortion,  etc.  To  meet  the  challenges  posed  by  Deepfake  multimedia,  we  propose  a  deep  ensemble    learning    technique    called    DeepfakeStack    for    detecting  such  manipulated  videos.  The  proposed  technique  combines a series of DL based state-of-art classification models and  creates  an  improved  composite  classifier.  Based  on  our  experiments, it is shown that DeepfakeStack outperforms other classifiers by achieving an accuracy of 99.65% and AUROC of 1.0   score   in   detecting   Deepfake.   Therefore,   our   method   provides   a   solid   basis   for   building   a   Realtime   Deepfake   detector. 

Abstract—Android is a most widespread mobile-based operating system having more than billions of active users with a high market impression that stimulated the cyber criminals to impulsion the malware into this operating system. During a couple of recent years, wide-ranging researches are conducted in the domain of malware analysis and detection in Android devices while Android already have implemented various security controls to solve the problems includes unique user ID (UID) for each application, system permissions, and its distribution platform Google Play. In this paper, we optimize and evaluate different types of machine learning algorithms by implementing a static analysis training different classifier in order to separate malware from non-malware applications running on Android OS. In our evaluation, we use 11,120 applications where 5,560 malware samples and 5,560 benign samples of DREBIN dataset and the accuracy we get more than 94% that will help to detect an application is malicious or not. This experiment also leads to developing a real-time malware scanner to decide whether an Android app can be installed or executed on Android devices. 

Abstract—Sorting a set of unsorted items is a task that happens in computer programming while a computer program has to follow a sequence of precise directions to accomplish that task. In order to find things quickly by making extreme values easy to see, sorting algorithm refers to specifying a technique to arrange the data in a particular order or format where maximum of communal orders is in arithmetic or lexicographical order. A lot of sorting algorithms has already been developed and these algorithms have enhanced the performance in the factors including time and space complexity, stability, correctness, definiteness, finiteness, effectiveness, etc. A new approach has been proposed in this paper in sorting algorithm called MinFinder to overcome some of the downsides and performs better compared to some conventional algorithms in terms of stability, computational time, complexity analysis. 

Abstract—The hasty growth of technology and social media has carried momentous changes to humanoid communication. Facebook, the largest online social media in the last few years has more than 200 million active users where more than 3.5 billion minutes are spent on Facebook daily. Since the competence of Facebook is subject to mostly on the processing of the massive volume of data. The volume of data is increasing day to day as well as the number of inactive and fake users. In this paper, we propose a new model using distributed data-base concept for management of users and their activities. This proposed model helps to keep the system scalable, reliable, and faster and let the Facebook accessible from anywhere with high accessibility. 

Abstract—Malware detection plays a key role in Android device security due to the popularity of Android with billions of active users that encouraging cyber criminals to push the malware into this operating system. The growth of malware is now becoming a serious problem. Recently, extensive research has been conducted to detect malware on Android devices using machine learning based methods profoundly depending on domain knowledge for manually extracting malicious features. In this paper, we evaluate tree-based machine learning algorithms by Stacked Generalization concept for detecting malware on Android in conjunction with implementing a substring based method for training the algorithms. We perform experiments on 11,120 samples containing 5,560 malware samples and 5,560 benign samples provided by DREBIN dataset on 8 malware families. The evaluation results show how stacked generalization achieves 97.92% validation accuracy for malware detection on DREBIN dataset. 

Abstract—Unmanned aerial vehicles (UAVs) or drones serve a wide range of applications from surveillance to combat missions. UAVs carry, collect, or communicate sensitive information which becomes a target for the attacks. Securing the communication network between the operator and the UAV is therefore crucial. So far, the networks used in most UAV applications are static, which allows more time and opportunity for the adversary to perform cyber-attacks on the UAV. In this paper we propose to study Moving Target Defense (MTD) technique against cyber-attacks on the drones including wireless network encryption and intrusion detection system. MTD technique change the static nature of the systems to increase both the difficulty and the cost (effort, time, and resources) of mounting attacks. For illustration purpose, a well-known cyber attack is performed on a popular commercial drone and results are presented to show the network vulnerabilities, damages caused due to the attacks and defense techniques to prevent the attacks . 

Abstract—The goals and applications of inpainting are numerous, from the restoration of damaged paintings and photographs to the removal/replacement of selected object. Reconstruction of missing or damaged portions of images is an ancient practice used extensively in artwork restoration. Also known as inpainting or retouching, this activity consists of filling in the missing areas or modifying the damaged ones in a non-detectable way by an observer not familiar with the original images. Applications of image inpainting range from restoration of photographs, films and paintings, to removal of occlusions, such as text, subtitles, stamps and publicity from images. Inpainting, the technique of modifying an image in an undetectable form, is as ancient as art itself. 

Abstract—The advent of 21st century has brought various technological improvements across several areas. This movement has transformed the way material was used to be connected. The outcome is predominantly noticeable in the academic setting where educators and students are no more bound by time and space. Using current innovative tools students can learn in a more time-efficient way than eternally. As a result, they are untying from learning, and the rate of drop-out is increasing day by day. At present, a large number of educators are going for technology driven academic classrooms where they can switch the teaching and learning activities more productively. In this paper, we are proposing a model for building innovative education products which enables the teachers to engage students with different educational exercises and interactive activities using computer/smart devices. This Parent-Student-Teacher model will enable educators to create, organize and share their curricula, lesson plans and classroom materials; also, the teachers to share knowledge, exchange ideas and get feedback from other educators where teachers can give performance points to students in a real-time manner and share the performance and behavior of their students with the parents and administrators. It enables teachers, students and parents to collaborate on a single platform by interactive white boards where they can give lessons write notes and save their work with just the tip of finger. 

Abstract—Diagnostic ultrasound (US) is an important tool in today's sophisticated medical diagnostics. Nearly every medical discipline benefits itself from this relatively inexpensive method that provides a view of the inner organs of the human body without exposing the patient to any harmful radiations. Medical diagnostic images are usually corrupted by noise during their acquisition and most of the noise is speckle noise. To solve this problem, instead of using adaptive filters which are widely used, No-Local Means based filters have been used to de-noise the images. Ultrasound images of four organs such as Abdomen, Ortho, Liver, Kidney, Brest and Prostrate of a Human body have been used and applied comparative analysis study to find out the output. These images were taken from Siemens SONOLINE G60 S System and the output was compared by matrices like SNR, RMSE, PSNR IMGQ and SSIM. The significance and compared results were shown in a tabular format. 

Abstract—In the digital world, inpainting is the algorithm used to replace or reconstruct lost, corrupted, or deteriorated parts of image data. Of the various proposed inpainting methods, convolution methods are the simplest and most efficient. In this paper, an enhanced inpainting model based on convolution theorem is proposed for digital images that preserves the edge and effectively estimates the lost or damaged parts of an image. In the proposed algorithm, a mask image is created dynamically to detect the image area to inpaint where most of the algorithms detect the missing parts of the image manually. Studies confirm the simplicity and effectiveness of our method, which also produces results that are comparable to those produced using other methods. 

Abstract—The advent of synthetic media and Deepfakes is pushing us to face an uncomfortable truth: video and images are no longer accurate recordings of reality. Each and every digital communication channel that our society relies on, whether it be audio, video, image, or even text, is in danger of being manipulated with Deepfake. This technology has been widely used to propagate mis- and disinformation, inflame political tensions, defame opposition, or even used for blackmailing someone. To identify Deepfake, researchers have proposed a variety of deep learning (DL) approaches, as the technology has become more complex, making it more difficult. However, this research expands our previously stated methodology, which showed that classical machine learning (ML) approaches alone can achieve superior performance in detecting Deepfake. In the ML-based approach, the traditional procedures of feature development and feature selection are followed by training, experimenting, and testing of ML classifiers. With the ML technique, the model provides better understandability and interpretability while consuming less computing resources. In addition, this paper conducts an omnibus test, which is called ‘Analysis of Variance (ANOVA)’ under the null hypothesis in order to compare the performances of multiple ML models by using several statistical hypothesis testing frameworks. Finally, we present results on several Deepfake datasets that are obtained relatively fast with comparable or superior performance to the state-of-the-art DL-based methods: 99.84% accuracy on FaceFore-cics++, 99.38% accuracy on DFDC, 99.66% accuracy on VDFD, and 99.43% on Celeb-DF datasets. According to our findings, classic ML algorithms can be used to develop a system that effectively detects Deepfake. 

Abstract—In the Internet Age, the proliferation of information through online discourse has increased dramatically in recent years. The escalation of Internet usage has led to an increased spread of misinformation related to important, controversial topics such as the COVID pandemic. The spread of misinformation related to important, controversial topics has led to massive societal ramifications with the World Health Organization (WHO) even labeling the spread of this COVID-related misinformation as an “infodemic”. Due to this, an increased focus has been put on being able to understand, interpret, and detect this misinformation. While this research focus has led to the creation of multiple datasets for COVID misinformation detection, these current datasets emphasize the usage of primarily textual information for this purpose. Existing work, involving these datasets, has made limited use of the implicit visual contents in this regard and has not yet properly utilized the potential of the valuable information that can be derived from the images plus infographics of misinformation articles and social media posts. Therefore, it is necessary to create more explicitly multimodal datasets that account for both text and images to identify misinformation. To address this limitation, we perform a unique analysis of three different multimodal datasets on COVID misinformation, by specifically studying the images associated with the online websites listed by them, and by developing a preliminary taxonomy, based on our findings, to determine the appropriate path forward towards building a prospective holistically multimodal dataset. To our knowledge, this initial study is a first of its kind with visual cues in the context of multimodal datasets on COVID misinformation.