Towards a Machine Learning Driven Trust Management Heuristic for the Internet of Vehicles.

The rapid proliferation of the emerging yet promising notion of the Internet-of-Vehicles (IoV) has led to the development of a variety of conventional trust assessment schemes to tackle insider attackers. The primary reliance of these frameworks is on the accumulation of individual trust attributes. While aggregating these influential parameters, weights are often associated with each individual attribute to reflect its impact on the final trust score. It is of paramount importance that such weights be precise to lead to an accurate trust assessment. Moreover, the value of the minimum acceptable trust threshold employed for the identification of dishonest vehicles needs to be carefully defined to avoid delayed or erroneous detection. This paper employs an IoT data set from CRAWDAD by suitably transforming it into an IoV format. This data set encompasses information regarding 18,226 interactions among 76 nodes, both honest and dishonest. First, the influencing parameters (i.e., packet delivery ratio, familiarity, timeliness and interaction frequency) were computed, and two feature matrices were formed. The first matrix (FM1) takes into account all the pairwise individual parameters as individual features, whereas the second matrix (FM2) considers the average of all pairwise computations performed for each individual parameter as one feature. Subsequently, unsupervised learning is employed to achieve the ground truth prior to applying supervised machine learning algorithms for classification purposes. It is worth noting that Subspace KNN yielded a perfect precision, recall, and the F1-score equal to 1 for individual parametric scores, whereas Subspace Discriminant returned an ideal precision, recall, and the F1-score equal to 1 for mean parametric scores. It is also evident from extensive simulations that FM2 yielded more accurate classification results compared to FM1. Furthermore, decision boundaries among honest and dishonest vehicles have also been computed for respective feature matrices.

Machine learning research towards combating COVID-19: Virus detection, spread prevention, and medical assistance.

COVID-19 was first discovered in December 2019 and has continued to rapidly spread across countries worldwide infecting thousands and millions of people. The virus is deadly, and people who are suffering from prior illnesses or are older than the age of 60 are at a higher risk of mortality. Medicine and Healthcare industries have surged towards finding a cure, and different policies have been amended to mitigate the spread of the virus. While Machine Learning (ML) methods have been widely used in other domains, there is now a high demand for ML-aided diagnosis systems for screening, tracking, predicting the spread of COVID-19 and finding a cure against it. In this paper, we present a journey of what role ML has played so far in combating the virus, mainly looking at it from a screening, forecasting, and vaccine perspective. We present a comprehensive survey of the ML algorithms and models that can be used on this expedition and aid with battling the virus.

Adaptive Observer Based Fault Tolerant Control for Sensor and Actuator Faults in Wind Turbines.

The installed wind energy generation capacity has been increasing dramatically all over the world. However, most wind turbines are installed in hostile environments, where regular operation needs to be ensured by effective fault tolerant control methods. An adaptive observer-based fault tolerant control scheme is proposed in this article to address the sensor and actuator faults that usually occur on the core subsystems of wind turbines. The fast adaptive fault estimation (FAFE) algorithm is adopted in the adaptive observers to accurately and rapidly located the faults. Based on the states and faults estimated by the adaptive observers, the state feedback fault tolerant controllers are designed to stabilize the system and compensate for the faults. The gain matrices of the controllers are calculated by the pole placement method. Simulation results illustrate that the proposed fault tolerant control scheme with the FAFE algorithm stabilizes the faulty system effectively and performs better than the baseline on the benchmark model of wind turbines.

Deep learning for computer-aided abnormalities classification in digital mammogram: A data-centric perspective.

Breast cancer is the most common type of cancer in women, and early abnormality detection using mammography can significantly improve breast cancer survival rates. Diverse datasets are required to improve the training and validation of deep learning (DL) systems for autonomous breast cancer diagnosis. However, only a small number of mammography datasets are publicly available. This constraint has created challenges when comparing different DL models using the same dataset. The primary contribution of this study is the comprehensive description of a selection of currently available public mammography datasets. The information available on publicly accessible datasets is summarized and their usability reviewed to enable more effective models to be developed for breast cancer detection and to improve understanding of existing models trained using these datasets. This study aims to bridge the existing knowledge gap by offering researchers and practitioners a valuable resource to develop and assess DL models in breast cancer diagnosis.

Beyond CNNs: Exploiting Further Inherent Symmetries in Medical Image Segmentation.

Automatic tumor or lesion segmentation is a crucial step in medical image analysis for computer-aided diagnosis. Although the existing methods based on convolutional neural networks (CNNs) have achieved the state-of-the-art performance, many challenges still remain in medical tumor segmentation. This is because, although the human visual system can detect symmetries in 2-D images effectively, regular CNNs can only exploit translation invariance, overlooking further inherent symmetries existing in medical images, such as rotations and reflections. To solve this problem, we propose a novel group equivariant segmentation framework by encoding those inherent symmetries for learning more precise representations. First, kernel-based equivariant operations are devised on each orientation, which allows it to effectively address the gaps of learning symmetries in existing approaches. Then, to keep segmentation networks globally equivariant, we design distinctive group layers with layer-wise symmetry constraints. Finally, based on our novel framework, extensive experiments conducted on real-world clinical data demonstrate that a group equivariant Res-UNet (called GER-UNet) outperforms its regular CNN-based counterpart and the state-of-the-art segmentation methods in the tasks of hepatic tumor segmentation, COVID-19 lung infection segmentation, and retinal vessel detection. More importantly, the newly built GER-UNet also shows potential in reducing the sample complexity and the redundancy of filters, upgrading current segmentation CNNs, and delineating organs on other medical imaging modalities.

Perturb and optimize users' location privacy using geo-indistinguishability and location semantics.

Location-based services (LBS) are capable of providing location-based information retrieval, traffic navigation, entertainment services, emergency rescues, and several similar services primarily on the premise of the geographic location of users or mobile devices. However, in the process of introducing a new user experience, it is also easy to expose users' specific location which can result in more private information leakage. Hence, the protection of location privacy remains one of the critical issues of the location-based services. Moreover, the areas where humans work and live have different location semantics and sensitivities according to their different social functions. Although the privacy protection of a user's real location can be achieved by the perturbation algorithm, the attackers may employ the semantics information of the perturbed location to infer a user's real location semantics in an attempt to spy on a user's privacy to certain extent. In order to mitigate the above semantics inference attack, and further improve the quality of the location-based services, this paper hereby proposes a user side location perturbation and optimization algorithm based on geo-indistinguishability and location semantics. The perturbation area satisfying geo-indistinguishability is thus generated according to the planar Laplace mechanism and optimized by combining the semantics information and time characteristics of the location. The optimum perturbed location that is able to satisfy the minimum loss of location-based service quality is selected via a linear programming method, and can be employed to replace the real location of the user so as to prevent the leakage of the privacy. Experimental comparison of the actual road network and location semantics dataset manifests that the proposed method reduces approximately 37% perturbation distance in contrast to the other state-of-the-art methods, maintains considerably lower similarity of location semantics, and improves region counting query accuracy by a margin of around 40%.

Adversarial Robustness of Deep Reinforcement Learning Based Dynamic Recommender Systems.

Adversarial attacks, e.g., adversarial perturbations of the input and adversarial samples, pose significant challenges to machine learning and deep learning techniques, including interactive recommendation systems. The latent embedding space of those techniques makes adversarial attacks challenging to detect at an early stage. Recent advance in causality shows that counterfactual can also be considered one of the ways to generate the adversarial samples drawn from different distribution as the training samples. We propose to explore adversarial examples and attack agnostic detection on reinforcement learning (RL)-based interactive recommendation systems. We first craft different types of adversarial examples by adding perturbations to the input and intervening on the casual factors. Then, we augment recommendation systems by detecting potential attacks with a deep learning-based classifier based on the crafted data. Finally, we study the attack strength and frequency of adversarial examples and evaluate our model on standard datasets with multiple crafting methods. Our extensive experiments show that most adversarial attacks are effective, and both attack strength and attack frequency impact the attack performance. The strategically-timed attack achieves comparative attack performance with only 1/3 to 1/2 attack frequency. Besides, our white-box detector trained with one crafting method has the generalization ability over several other crafting methods.

Deep graph level anomaly detection with contrastive learning.

Graph level anomaly detection (GLAD) aims to spot anomalous graphs that structure pattern and feature information are different from most normal graphs in a graph set, which is rarely studied by other researchers but has significant application value. For instance, GLAD can be used to distinguish some different characteristic molecules in drug discovery and chemical analysis. However, GLAD mainly faces the following three challenges: (1) learning more comprehensive graph level representations to differ normal graphs and abnormal graphs, (2) designing an effective graph anomaly evaluation paradigm to capture graph anomalies from the local and global graph perspectives, (3) overcoming the number imbalance problem of normal and abnormal graphs. In this paper, we combine graph neural networks and contrastive learning to build an end-to-end GLAD framework for solving the three challenges above. We aim to design a new graph level anomaly evaluation way, which first utilizes the contrastive learning strategy to enhance different level representations of normal graphs from node and graph levels by a graph convolution autoencoder with perturbed graph encoder. Then, we evaluate the error of them with corresponding representations of the generated reconstruction graph to detect anomalous graphs. Extensive experiments on ten real-world datasets from three areas, such as molecular, protein and social network anomaly graphs, show that our model can effectively detect graph level anomaly from the majority and outperform existing advanced methods.

A Comprehensive Survey on Community Detection With Deep Learning.

Detecting a community in a network is a matter of discerning the distinct features and connections of a group of members that are different from those in other communities. The ability to do this is of great significance in network analysis. However, beyond the classic spectral clustering and statistical inference methods, there have been significant developments with deep learning techniques for community detection in recent years--particularly when it comes to handling high-dimensional network data. Hence, a comprehensive review of the latest progress in community detection through deep learning is timely. To frame the survey, we have devised a new taxonomy covering different state-of-the-art methods, including deep learning models based on deep neural networks (DNNs), deep nonnegative matrix factorization, and deep sparse filtering. The main category, i.e., DNNs, is further divided into convolutional networks, graph attention networks, generative adversarial networks, and autoencoders. The popular benchmark datasets, evaluation metrics, and open-source implementations to address experimentation settings are also summarized. This is followed by a discussion on the practical applications of community detection in various domains. The survey concludes with suggestions of challenging topics that would make for fruitful future research directions in this fast-growing deep learning field.

Dataset for toothbrushing activity using brush-attached and wearable sensors.

Maintaining oral hygiene is very important for a healthy life. Poor toothbrushing is one of the leading causes of tooth decay and other gum problems. Many people do not brush their teeth properly. There is very limited technology available to help in assessing the quality of toothbrushing. Human Activity Recognition (HAR) applications have seen a tremendous growth in recent years. In this work, we treat the adherence to standard toothbrushing practice as an activity recognition problem. We investigate this problem and collect experimental data using a brush-attached and a wearable sensor when the users brush their teeth. In this paper, we extend our previous dataset [1] for toothbrushing activity by including more experiments and adding a new sensor. We discuss and analyse the collection of the dataset. We use an Inertial Measurement Unit (IMU) sensor to collect the time-series data for toothbrushing activity. We recruited 22 healthy participants and collected the data in two different settings when they brushed their teeth in five different locations using both electric and manual brushes. In total, we have recorded 120 toothbrushing sessions using both brush-attached sensor and the wearable sensor.

Nonrigid Point Set Registration With Robust Transformation Learning Under Manifold Regularization.

This paper solves the problem of nonrigid point set registration by designing a robust transformation learning scheme. The principle is to iteratively establish point correspondences and learn the nonrigid transformation between two given sets of points. In particular, the local feature descriptors are used to search the correspondences and some unknown outliers will be inevitably introduced. To precisely learn the underlying transformation from noisy correspondences, we cast the point set registration into a semisupervised learning problem, where a set of indicator variables is adopted to help distinguish outliers in a mixture model. To exploit the intrinsic structure of a point set, we constrain the transformation with manifold regularization which plays a role of prior knowledge. Moreover, the transformation is modeled in the reproducing kernel Hilbert space, and a sparsity-induced approximation is utilized to boost efficiency. We apply the proposed method to learning motion flows between image pairs of similar scenes for visual homing, which is a specific type of mobile robot navigation. Extensive experiments on several publicly available data sets reveal the superiority of the proposed method over state-of-the-art competitors, particularly in the context of the degenerated data.