In a personalized product recommendation system, when the amount of log data is large or sparse, the accuracy of model recommendation will be greatly affected. To solve this problem, a personalized product recommendation method using deep factorization machine (DeepFM) to analyze user behavior is proposed. Firstly, the K-means clustering algorithm is used to cluster the original log data from the perspective of similarity to reduce the data dimension. Then, through the DeepFM parameter sharing strategy, the relationship between low- and high-order feature combinations is learned from log data, and the click rate prediction model is constructed. Finally, based on the predicted click-through rate, products are recommended to users in sequence and fed back. The area under the curve (AUC) and Logloss of the proposed method are 0.8834 and 0.0253, respectively, on the Criteo dataset, and 0.7836 and 0.0348 on the KDD2012 Cup dataset, respectively. Compared with other newer recommendation methods, the proposed method can achieve better recommendation effect.

http://www.koreascience.or.kr:80/article/JAKO202116057062097.pdf

Personalized Product Recommendation Method for Analyzing User Behavior Using DeepFM

With the exponential growth of connected Internet of Things (IoT) devices around the world, security protection and privacy preservation have risen to the forefront of design and development of innovative systems and services. For low-value IoT devices that identify and track billion of goods in various industries—such as radio-frequency identification (RFID) tags—this involves multiple challenges in very constrained environments. IoT devices aim to design low-cost, low-complexity infrastructure while enabling robust authentication protocols with reduced latency and energy consumption. Given these challenges, in this article, we present a new lightweight authentication protocol for IoT applications, employing an authenticated-encryption (AE) cryptosystem with associated data (AEAD). Since AEAD algorithms provide data confidentiality and message integrity simultaneously, security analysis [Real-or-Random (RoR) and Scyther] results prove the robustness of the proposed protocol against IoT threats. Furthermore, to measure the computation and communication cost, FPGA and ASIC simulations using four different AEAD candidates of National Institute of Standards and Technology (NIST) lightweight cryptography competition are executed. The implementation results [e.g., 4744 gate equivalent (GE) and 0.87-mw power] clearly show that our novel design can be applied to a wide range of constrained IoT devices complying with low-cost, lightweight, and high-speed requirements.

An Authentication Protocol for Next Generation of Constrained IoT Systems

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10028215

High potential of technology to face new respiratory viruses: mechanical ventilation devices for effective healthcare to next pandemic emergencies

Objective. In the present work, the bibliographical production referring to smart contracts is reviewed from the business, academic and scientific points of view. The aim is to analyze precisely how the relationship between smart contracts and blockchain occurs to establish a context in both developments. 
Design/Methodology/Approach. The Scopus database was consulted using the keyword: "smart contract". It was decided to use the SciMAT bibliometric tool for the bibliometric study with the approach based on a joint analysis of words and the h-index to visualize diagrams and thematic areas relatively simply for two periods: 2010-2018 and 2019-2021. 
Results/Discussion. It was necessary to know the concept of smart contracts for the future of contract law and the digital economy. Some institutions and people will be against considering them as authorized legal alternatives; However, due to the technological advance currently being presented, their development will only be a matter of time. 
Conclusions. Smart contracts are a technological innovation that, through blockchain technology, carry out transactions safely, quickly and with almost no risk of non-compliance.

/pdf/opportunities-and-applications-of-smart-contracts-a-vision-1dkdgpz3.pdf

Opportunities and applications of smart contracts: A vision from the business, academic and scientific literature

Technology to face pandemic crises: Mechanical ventilators improve health emergency preparedness

Increasingly, there have been attempts to utilize physiological information collected from different non-intrusive devices and sensors (e.g., electroencephalogram, electrocardiograph, electrodermal activity, and skin conductance) for different activities and studies, such as using the data to train machine/deep-learning models in order to facilitate medical diagnosis and other decision-making. Given the constant advances in machine and deep learning methods, such as deep learning, transfer learning, reinforcement learning, and federated learning, we can also utilize such techniques in cognitive computing to facilitate human behavior analysis. For example, transfer learning uses data or knowledge acquired on solved problems to help solve unsolved but very relevant problems. Transfer learning is often used in cognitive computing to use differences between individuals or tasks to improve learning efficiency and effectiveness. Transfer learning can also be integrated with deep learning to take advantage of the progress of deep learning and transfer learning. However, there are still many major challenges when using deep and machine learning algorithms to analyze human behavior. For example, deep and machine learning is used to learn the feature representation of human behavior from multimodal data; deep and machine learning is used to realize pattern mapping from one modal to another; deep and machine learning is used to realize the fusion of multiple modal data to achieve more high prediction accuracy; deep and machine learning is used to restore missing data in the modal, etc. Not surprisingly, several multimodal machine learning models have been developed in recent years, which have shown promising results when applied on applications such as multimedia descriptions and retrieval. Therefore, we posit the potential of leveraging such advances to address fundamental challenges in human behavior analysis. In this thematic issue, we seek to provide a forum for researchers from cognitive computing and machine learning to present recent progress in deep and advanced machine learning research with applications to multimodal human behavior data. According to peer reviewers’ review comments, this special issue contains a total of seven papers, and the authors have revised the papers based on the review comments [1-7]. This special issue has been possible by the strong support of Prof. James Jong Hyuk Park, the Editor-in-Chief of Journal of Information Processing Systems (JIPS). We would like to express our gratitude to the authors for their contributions.

http://www.koreascience.or.kr:80/article/JAKO202116056956032.pdf

Kim-Kwang Raymond Choo

Papers

A Special Section on Deep & Advanced Machine Learning Approaches for Human Behavior Analysis

Internet of things: Conceptual network structure, main challenges and future directions

Resilience capabilities of healthcare supply chain and supportive digital technologies

2DF-IDS: Decentralized and differentially private federated learning-based intrusion detection system for industrial IoT

Network traffic classification: Techniques, datasets, and challenges