Recent Advances in Computer Science and Communications - Volume 16, Issue 7, 2023

In today’s era, Deep Learning has become the center of recent ascent in the field of artificial intelligence and its models. There are various Artificial Intelligence models that can be viewed as needing more strength for adversely defined information sources. It also leads to a high potential security concern in the adversarial paradigm; the DNN can also misclassify inputs that appear to expect in the result. DNN can solve complex problems accurately. It is empaneled in the vision research area to learn deep neural models for many tasks involving critical security applications. We have also revisited the contributions of computer vision in adversarial attacks on deep learning and discussed its defenses. Many of the authors have given new ideas in this area, which has evolved significantly since witnessing the first-generation methods. For optimal correctness of various research and authenticity, the focus is on peer-reviewed articles issued in the prestigious sources of computer vision and deep learning. Apart from the literature review, this paper defines some standard technical terms for non-experts in the field. This paper represents the review of the adversarial attacks via various methods and techniques along with their defenses within the deep learning area and future scope. Lastly, we bring out the survey to provide a viewpoint of the research in this Computer Vision area.

Background: As innovative information technology, blockchain has combined the advantages of decentralization, immutability, data provenance, and contract operation automatically, which can be used to solve the issues of single point failure, high trading cost, low effectiveness, and data potential risk in power trading. However, in the traditional power blockchain, the design of functional components in blockchain, such as the data structure of the block, does not take the actual features of power into account, thus leading to a performance bottleneck in practical application. Motivated by business characteristics of power trading, a user-centric data model UCDM in consortium blockchain is proposed to achieve efficient data storage and quick data retrieval. Methods: The proposed UCDM is designed by considering the requirements of transaction retrieval and analysis, thus supporting the requirements of concurrent data requests and mass data storage. The ID of each user will independently form its own chain over the blockchain. Results: Compared with the traditional data model, the extensive experimental results demonstrate that the proposed UCDM has shorter processing delay, higher throughput, and shorter response latency, thus having practical value. Conclusion: UCDM is an effective solution to the transaction retrieval and analysis in power blockchain. Furthermore, the participant of the blockchain network has a unique identity over the world, which ensures high security during trading.

Objective: Open-source software (OSS) has become an important choice for developing software applications, and its usage has exponentially increased in recent years. Although many OSS systems have shown high reliability in terms of their functionality, they often exhibit several quality issues. Since most developers focus primarily on meeting clients’ functional requirements within the appropriate deadlines, the outcome suffers from poor design and implementation practices. This issue can also manifest as software code smells, resulting in a variety of quality issues such as software maintainability, comprehensibility, and extensibility. Generally speaking, OSS developers use code reviews during their software development to discover flaws or bugs in the updated code before it is merged with the code base. Nevertheless, despite the harmful impacts of code smells on software projects, the extent to which developers do consider them in the code review process is unclear in practice. Methods: To better understand the code review process in OSS projects, we gathered the comments of code reviewers who specified where developers should fix code smells in two OSS projects, OpenStack and WikiMedia, between 2011 and 2015. Results: Our findings indicate that most code reviewers do not pay much attention to code smells. Only a few code reviewers have attempted to motivate developers to improve their source code quality in general. The results also show that there is an increasing tendency to provide advice concerning code smells corrections over time. Conclusion: We believe that this study's findings will encourage developers to use new software engineering practices, such as refactoring, to reduce code smells when developing OSS.

Objective: As GAN-based deepfakes have become increasingly mature and realistic, the demand for effective deepfake detectors has become essential. We are inspired by the fact that normal pulse rhythms present in real-face video can be decreased or even completely interrupted in a deepfake video; thus, we have introduced a new deepfake detection approach based on remote heart rate estimation using the 3D Central Difference Convolution Attention Network (CDCAN). Methods: Our proposed fake detector is mainly composed of a 3D CDCAN with an inverse attention mechanism and LSTM architecture. It utilizes 3D central difference convolution to enhance the spatiotemporal representation, which can capture rich physiological-related temporal context by gathering the time difference information. The soft attention mechanism is to focus on the skin region of interest, while the inverse attention mechanism is to further denoise rPPG signals. Results: The performance of our approach is evaluated on the two latest Celeb-DF and DFDC datasets, for which the experiment results show that our proposed approach achieves an accuracy of 99.5% and 97.4%, respectively. Conclusion: Our approach outperforms the state-of-art methods and proves the effectiveness of our DeepFake detector.

Background: The image generation model based on generative adversarial network (GAN) has achieved remarkable achievements. However, traditional GAN has the disadvantage of unstable training, which affects the quality of the generated image. Objective: This method is to solve the GAN image generation problems of poor image quality, single image category, and slow model convergence. Methods: An improved image generation method is proposed based on GAN. Firstly, the attention mechanism is introduced into the convolution layer of the generator and discriminator and a batch normalization layer is added after each convolution layer. Secondly, the ReLU and leaky ReLU are used as the active layer of the generator and discriminator, respectively. Thirdly, the transposed convolution is used in the generator while the small step convolution is used in the discriminator, respectively. Fourthly, a new discarding method is applied in the dropout layer. Results: The experiments were carried out on Caltech 101 dataset. The experimental results showed that the image quality generated by the proposed method is better than that generated by GAN with attention mechanism (AM-GAN) and GAN with stable training strategy (STS-GAN) and the stability was improved. Conclusion: The proposed method is effectiveness for image generation with high quality.

In typical Internet of Things (IoT) networks, data is sent from sensors to fog devices and then onto a central cloud server. One single point of failure, a slowdown in the flow of data, identification, security, connection, privacy concerns caused by a third party managing cloud servers, and the difficulty of frequently updating the firmware on millions of smart devices from both a maintenance and a security point of view are just some of the problems that can occur. The evolution of ubiquitous computing and blockchain technology has inspired researchers worldwide in recent years. Key features of blockchain technology, such as the fact that it can't be changed and a decentralised and distributed approach to data security, have made it a popular choice for developing diverse applications. With the practically significant applicability of blockchain concepts (specifically consensus methods), modern-day applications in ubiquitous computing and other related areas have significantly benefited. In addition, we have taken advantage of the widely available blockchain platforms and looked into potential new study fields. As a result, this review paper elaborates the novel alternative privacy preservation options while simultaneously focusing on the universal domain as a starting point for blockchain technology applications. We also discuss obstacles, research gaps, and solutions. This review can assist early researchers who are beginning to investigate the applicability of blockchain technology in ubiquitous computing. It is also possible to use it as a reference in order to speed up the process of finding the appropriate markers for ongoing research subjects that are of interest.

Introduction: Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) both have their areas of specialty in the medical imaging world. MRI is considered to be a safer modality as it exploits the magnetic properties of the hydrogen nucleus. Whereas a CT scan uses multiple X-rays, which is known to contribute to carcinogenesis and is associated with affecting the patient's health. Methods: In scenarios, such as radiation therapy, where both MRI and CT are required for medical treatment, a unique approach to getting both scans would be to obtain MRI and generate a CT scan from it. Current deep learning methods for MRI to CT synthesis purely use either paired data or unpaired data. Models trained with paired data suffer due to a lack of availability of wellaligned data. Results: Training with unpaired data might generate visually realistic images, although it still does not guarantee good accuracy. To overcome this, we proposed a new model called PUPCGANs (Paired Unpaired CycleGANs), based on CycleGANs (Cycle-Consistent Adversarial Networks). Conclusion: This model is capable of learning transformations utilizing both paired and unpaired data. To support this, a paired loss is introduced. Comparing MAE, MSE, NRMSE, PSNR, and SSIM metrics, PUPC-GANs outperform CycleGANs.

Background: The watermarking technique is a security algorithm for medical images and the patient's information. Watermarking is used for maintaining the robustness, integrity, confidentiality, authentication, and complexity of medical images. Objective: The selection of medical image watermarking technique is multi-criteria decisionmaking and an automatic way of algorithm selection for security and privacy. However, it is difficult to select a better watermarking technique through traditional selection techniques. Methods: To deal with this problem, a multicriteria-based fuzzy analytic hierarchy process (FAHP) was proposed. This method is applied for algorithm selection for the security of medical images in healthcare. In this method, we first determined the list of criteria and alternatives, which directly affect the decision of the medical image security algorithm. Then, the proposed method was applied to the criteria and alternatives. Results: We provided the rank according to the obtained weights of the algorithm. Furthermore, the overall results and ranking of the algorithms were also presented in this article. Conclusion: Integrity was found to have the highest weight (0.509) compared to the other criteria. The weight of the other criteria, namely authentication, was 0.165, robustness was 0.151, confidentiality was 0.135, and complexity was 0.038. Thus, in terms of ranking, integrity was reported to be of the highest priority among the five criteria attributes.

Introduction: Deep neural networks due to their linear nature are sensitive to adversarial examples. They can easily be broken just by a small disturbance to the input data. Some of the existing methods to perform these kinds of attacks are pixel-level perturbation and spatial transformation of images. Method: These methods generate adversarial examples that can be fed to the network for wrong predictions. The drawback that comes with these methods is that they are really slow and computationally expensive. This research work performed a black box attack on the target model classifier by using the generative adversarial networks (GAN) to generate adversarial examples that can fool a classifier model to classify the images as wrong classes. The proposed method used a biased dataset that does not contain any data of the target label to train the first generator Gnorm of the first stage GAN, and after the first training has finished, the second stage generator Gadv, which is a new generator model that does not take random noise as input but the output of the first generator Gnorm. Result: The generated examples have been superimposed with the Gnorm output with a small constant, and then the superimposed data have been fed to the target model classifier to calculate the loss. Some additional losses have been included to constrain the generation from generating target examples. Conclusion: The proposed model has shown a better fidelity score, as evaluated using Fretchet inception distance score (FID), which was up to 42.43 in the first stage and up to 105.65 in the second stage with the attack success rate of up to 99.13%.