Currently, many applications have emerged from the implementation of software development and hardware use, known as the Internet of things. One of the most important application areas of this type of technology is in health care. Various applications arise daily in order to improve the quality of life and to promote an improvement in the treatments of patients at home that suffer from different pathologies. That is why there has emerged a line of work of great interest, focused on the study and analysis of daily life activities, on the use of different data analysis techniques to identify and to help manage this type of patient. This article shows the result of the systematic review of the literature on the use of the Clustering method, which is one of the most used techniques in the analysis of unsupervised data applied to activities of daily living, as well as the description of variables of high importance as a year of publication, type of article, most used algorithms, types of dataset used, and metrics implemented. These data will allow the reader to locate the recent results of the application of this technique to a particular area of knowledge.

https://www.mdpi.com/1424-8220/20/9/2702/pdf

Unsupervised Human Activity Recognition Using the Clustering Approach: A Review

Wireless sensor network (WSN), which serves as one of the key components of the cyber physical system, is a wireless network composed of many stationary or moving sensors in a self-organizing and multi-hop manner. It cooperatively senses, collects, processes and transmits the information of the perceived objects in the geographical area covered by the network, and finally sends this information to the owner of the network. There are some typical attacks in WSN, such as Blackhole, Grayhole, Flooding and Scheduling attacks, which may cause damages to the WSN system in a short time. Moreover, intrusion detection methods for WSN suffer from the disadvantages of low detection rate, large calculation overhead, and high false alarm rate due to the limited resources of sensor nodes and a large amount of redundancy as well as high correlation of network data. In view of the above problems, we propose SLGBM, an intrusion detection method for wireless sensor networks. First, the sequence backward selection (SBS) algorithm is applied to reduce the data dimension on the feature space of the original traffic data so as to reduce the computational overhead. A LightGBM algorithm is then utilized to detect different network attacks. The experimental results based on the WSN-DS dataset show that F-measure of our proposed SLGBM are obviously superior to current typical detection methods, which are 99.8%, 99.4%, 99.1%, 96.5%, and 96.1% in Normal, Blackhole, Grayhole, Flooding, and Scheduling (TDMA) attack detections respectively.

/pdf/slgbm-an-intrusion-detection-mechanism-for-wireless-sensor-3irkklkmu5.pdf

SLGBM: An Intrusion Detection Mechanism for Wireless Sensor Networks in Smart Environments

As we make tremendous advances in machine learning and artificial intelligence technosciences, there is a renewed understanding in the AI community that we must ensure that humans being are at the center of our deliberations so that we don't end in technology-induced dystopias. As strongly argued by Green in his book Smart Enough City, the incorporation of technology in city environs does not automatically translate into prosperity, wellbeing, urban livability, or social justice. There is a great need to deliberate on the future of the cities worth living and designing. There are philosophical and ethical questions involved along with various challenges that relate to the security, safety, and interpretability of AI algorithms that will form the technological bedrock of future cities. Several research institutes on human centered AI have been established at top international universities. Globally there are calls for technology to be made more humane and human-compatible. For example, Stuart Russell has a book called Human Compatible AI. The Center for Humane Technology advocates for regulators and technology companies to avoid business models and product features that contribute to social problems such as extremism, polarization, misinformation, and Internet addiction. In this paper, we analyze and explore key challenges including security, robustness, interpretability, and ethical challenges to a successful deployment of AI or ML in human-centric applications, with a particular emphasis on the convergence of these challenges. We provide a detailed review of existing literature on these key challenges and analyze how one of these challenges may lead to others or help in solving other challenges. The paper also advises on the current limitations, pitfalls, and future directions of research in these domains, and how it can fill the current gaps and lead to better solutions.

Developing Future Human-Centered Smart Cities: Critical Analysis of Smart City Security, Interpretability, and Ethical Challenges.

https://www.techscience.com/iasc/v26n6/41010/pdf

A Pursuit of Sustainable Privacy Protection in Big Data Environment by an Optimized Clustered-Purpose Based Algorithm

The application of composite materials in vehicle thin-walled structure has attracted more and more interest by carmakers. This study aimed to investigate into the dynamic impact responses of the carbon fiber reinforced plastic (CFRP) header rail subjected to the vertical (Z-direction) and horizontal direction (X-direction) loading. The lightweight CFRP header rail was first fabricated using the wet compression molding process, special fixtures were designed and prepared for drop weight impact tests. The failure modes, force responses, energy absorption characteristics and the structural responses to different directional loads were compared. The global peak forces in X-directional tests were higher than those in Z-directional counterpart, but the energy absorption ratios were lower. It is shown that the impact performance of the specimens under Z-directional impact was better than under X-directional impact with same loading magnitude. The finite element (FE) models were then established in both impact directions for exploring the failure mechanisms. The simulated impact process, force–displacement curves and failure modes were found to be consistent with the experimental counterparts. After validation, damaged and weak areas of the specimen could be identified from the simulations, and the models could effectively predict the global peak force and energy absorption ratio. • Dynamic responses of CFRP header rail were investigated under vertical and horizontal impact. • The impact performance of the specimen was better under vertical impact than horizontal impact. • Failure modes and main contributors to energy absorption between different tests were compared. • Damaged and weak areas of the specimen were identified from the FE simulations. 

Experimental and numerical investigation into the dynamic impact responses of CFRP header rail

Numerical investigation into impact responses of an offshore wind turbine jacket foundation subjected to ship collision

/pdf/defense-against-poisoning-attack-via-evaluating-training-dqg11ioc5e.pdf

Defense against poisoning attack via evaluating training samples using multiple spectral clustering aggregation method

Recently, machine learning-based wireless intrusion detection systems (IDSs) have been demonstrated to have high detection accuracy in malicious traffic detection. However, many researchers argue that a variety of attacks are significantly challenging the security of machine learning techniques themselves. In this paper, we study two different types of security threats which can effectively degrade the performance of machine learning based wireless IDSs. First, we propose an Adaptive SMOTE (A-SMOTE) algorithm which can adaptively generate new training data points based on few existing ones with labels. Then, we introduce a stealing model attack by training a substitute model using deep neural networks (DNNs) based on the augmented training data in order to imitate the machine learning model embedded in targeted systems. After that, we present a novel poisoning strategy to attack against the substitute machine learning model, resulting in a set of adversarial samples that can be used to degrade the performance of targeted systems. Experiments on three real data sets collected from wired and wireless networks have demonstrated that the proposed stealing model and poisoning attacks can effectively degrade the performance of IDSs using different machine learning algorithms.

Poisoning Machine Learning Based Wireless IDSs via Stealing Learning Model

Machine learning models are vulnerable to a variety of data perturbation. Recent research mainly focuses on the vulnerability of model training and proposes various model-oriented defense methods to achieve robust machine learning. However, most of the existing research overlooks the vulnerability of model capacity, which is more fundamental for model performance. In this paper, we study an adversarial vulnerability of model capacity caused by the poisoning on the estimation of model hyper-parameters. We further implement this vulnerability catering for the polynomial regression model, on which the evading of model-oriented detection is challenging, to illustrate the effectiveness of the adversarial vulnerability. Extensive experiments on one synthetic and three real-world data sets demonstrate that the vulnerability can effectively mislead the hyper-parameter estimation of the polynomial regression model by poisoning a few numbers of camouflage samples that cannot be detected by model-oriented defense methods.

/pdf/model-capacity-vulnerability-in-hyper-parameters-estimation-453stzfdqx.pdf

Model Capacity Vulnerability in Hyper-Parameters Estimation

Adding perturbations via utilizing auxiliary gradient information or discarding existing details of the benign images are two common approaches for generating adversarial examples. Though visual imperceptibility is the desired property of adversarial examples, conventional adversarial attacks still generate traceable adversarial perturbations. In this paper, we introduce a novel Adversarial Attack via Invertible Neural Networks (AdvINN) method to produce robust and imperceptible adversarial examples. Specifically, AdvINN fully takes advantage of the information preservation property of Invertible Neural Networks and thereby generates adversarial examples by simultaneously adding class-specific semantic information of the target class and dropping discriminant information of the original class. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet-1K demonstrate that the proposed AdvINN method can produce less imperceptible adversarial images than the state-of-the-art methods and AdvINN yields more robust adversarial examples with high confidence compared to other adversarial attacks. Code is available at https://github.com/jjhuangcs/AdvINN.

Xiao Liu

Papers

Numerical investigation into impact responses of an offshore wind turbine jacket foundation subjected to ship collision

Defense against poisoning attack via evaluating training samples using multiple spectral clustering aggregation method

Poisoning Machine Learning Based Wireless IDSs via Stealing Learning Model

Model Capacity Vulnerability in Hyper-Parameters Estimation

Imperceptible Adversarial Attack via Invertible Neural Networks