Search

Here we report on the issue of Advanced Persistent Threats (APT), which use malware for the purpose of leaking the data of large corporations and government agencies. APT attacks target systems continuously by utilizing intelligent and complex technologies. To overthrow the elaborate security network of target systems, it conducts an attack after undergoing a pre-reconnaissance phase. An APT attack causes financial loss, information leakage, etc. They can easily bypass the antivirus system of a target system. In this paper, we propose a Multi-Layer Defense System (MLDS) that can defend against APT. This system applies a reinforced defense system by collecting and analyzing log information and various information from devices, by installing the agent on the network appliance, server and end-user. It also discusses how to detect an APT attack when one cannot block the initial intrusion while continuing to conduct other activities. Thus, this system is able to minimize the possibility of initial intrusion and damages of the system by promptly responding through rapid detection of an attack when the target system is attacked.

Recently, intelligent transport systems have been applied to vehicle cloud environments. Such technology is especially useful for the systematic management of road traffic. Moreover, automobiles are increasingly equipped with a black box for accident prevention and preservation of evidence. Vehicle black boxes have become mandatory because black box images and voice data have served as forensic evidence in courts. However, the data from black boxes can be forged or modified by man-in-the-middle (MITM) attacks and message hijacking. In this paper, we propose a vehicle cloud computing-based black box service model that can provide integrity for black box data through digital signatures in vehicle cloud computing (VCC) environments. Our proposed model protects against MITM attacks and message hijacking using only a hash value and digital signature. Moreover, a mirroring technique (RAID 1) provides backup and recovery to protect the data from a traffic accident.

This paper proposes a hybrid deep learning emotion classification system (HDECS), a hybrid multimodal deep learning system designed for emotion classification in a specific national language. Emotion classification is important in diverse fields, including tailored corporate services, AI advancement, and more. Additionally, most sentiment classification techniques in speaking situations are based on a single modality: voice, conversational text, vital signs, etc. However, analyzing these data presents challenges because of the variations in vocal intonation, text structures, and the impact of external stimuli on physiological signals. Korean poses challenges in natural language processing, including subject omission and spacing issues. To overcome these challenges and enhance emotion classification performance, this paper presents a case study using Korean multimodal data. The case study model involves retraining two pretrained models, LSTM and CNN, until their predictions on the entire dataset reach an agreement rate exceeding 0.75. Predictions are used to generate emotional sentences appended to script data, which are further processed using BERT for final emotion prediction. The research result is evaluated by using categorical cross-entropy (CCE) to measure the difference between the model’s predictions and actual labels, F1 score, and accuracy. According to the evaluation, the case model outperforms the existing KLUE/roBERTa model with improvements of 0.5 in CCE, 0.09 in accuracy, and 0.11 in F1 score. As a result, the HDECS is expected to perform well not only on Korean multimodal datasets but also on sentiment classification considering the speech characteristics of various languages and regions.

This paper proposes a recommendation system based on a hybrid learning approach for a personal deep sleep service, called the Customized Deep Sleep Recommender System (CDSRS). Sleep is one of the most important factors for human life in modern society. Optimal sleep contributes to increasing work efficiency and controlling overall well-being. Therefore, a sleep recommendation service is considered a necessary service for modern individuals. Accurate sleep analysis and data are required to provide such a personalized sleep service. However, given the variations in sleep patterns between individuals, there is currently no international standard for sleep. Additionally, service platforms face a cold start problem when dealing with new users. To address these challenges, this study utilizes K-means clustering analysis to define sleep patterns and employs a hybrid learning algorithm to evaluate recommendations by combining user-based and collaborative filtering methods. It also incorporates feedback top-N classification processing for user profile learning and recommendations. The behavior of the study model is as follows. Using personal information received through mobile devices and data, such as snoring, sleep time, movement, and noise collected through AI motion beds, we recommend sleep and receive user evaluations of recommended sleep. This assessment reconstructs the profile and, finally, makes recommendations using top-N classification. The experimental results were evaluated using two absolute error measurement methods: mean squared error (MSE) and mean absolute percentage error (MAPE). The research results regarding the hybrid learning methods show 13.2% fewer errors than collaborative filtering (CF) and 10.2% fewer errors than content-based filtering (CBF) on an MSE basis. According to the MAPE, the methods are 14.7% more accurate than the CF model and 9.2% better than the CBF model. These results demonstrate that CDSRS systems can provide more accurate recommendations and customized sleep services to users than CF, CBF, and combination models. As a result, CDSRS, a hybrid learning method, can better reflect a user’s evaluation than traditional methods and can increase the accuracy of recommendations as the number of users increases.

Mangiferin is a natural immunomodulator found in plants including mango trees. The effects of mangiferin on chondrogenesis and cartilage repair have not yet been reported. This study was designed to determine the effect of mangiferin on chondrogenic differentiation in IL-1β-stimulated mesenchymal stem cells (MSCs) from subchondral bone and to explore the mechanisms underlying these effects. MSCs were isolated from the subchondral bone of rabbit and treated with mangiferin alone and/or interleukin-1β (IL-1β). Mangiferin induced chondrogenic differentiation in MSCs by upregulating transforming growth factor (TGF)-β, bone morphogenetic protein (BMP)-2, and BMP-4 and several key markers of chondrogenesis, including sex-determining region Y–box (SRY-box) containing gene 9 (SOX9), type 2α1 collagen (Col2α1), cartilage link protein, and aggrecan. In IL-1β-stimulated MSCs, mangiferin significantly reversed the production of TGF-β, BMP-2, BMP-4, SOX9, Col2α1, cartilage link protein, and aggrecan, as well as matrix metalloproteinase (MMP)-1, MMP-13, and a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS5). Mangiferin upregulated the phosphorylation of Smad 2, Smad 3, Smad 1/5/8, and SOX9 in IL-1β-stimulated MSCs. In the presence of mangiferin, SOX9 siRNA suppressed the activation of Smad 2, Smad 3, Smad 1/5/8, aggrecan, and Col2α1 expression. In conclusion, mangiferin exhibits both chondrogenic and chondroprotective effects on damaged MSCs and mediates these effects by targeting multiple aspects of the Smad and SOX9 signaling pathways.

Arginine, an α-amino acid, has been reported to exert beneficial effects that ameliorate health problems and prevent excessive fat deposition. In this study, we investigated whether the activation of cell signaling by arginine can induce osteogenic differentiation and modulate excessive adipogenic differentiation in human mesenchymal stem cells (MSCs). Arginine potently induced the expression of type Iα1 collagen, osteocalcin, and ALP in a dose-dependent manner without causing cytotoxicity. Arginine significantly increased the mRNA expression of the osteogenic transcription factors runt-related transcription factor 2 (Runx2), DIx5, and osterix. Furthermore, arginine demonstrated its antiadipogenicity by decreasing adipocyte formation and triglyceride (TG) content in MSCs and inhibiting the mRNA expression of the adipogenic transcription factors peroxisome proliferator-activated receptor γ (PPARγ), CCAAT/enhancer-binding protein α (C/EBPα), and fatty acid binding protein 4 (Fabp4). This effect was associated with increased expression of Wnt5a, and nuclear factor of activated T-cells (NFATc), and was abrogated by antagonists of Wnt and NFATc, which indicated a role of Wnt and NFATc signaling in the switch from adipogenesis to osteoblastogenesis induced by arginine. In conclusion, this is the first report of the dual action of arginine in promoting osteogenesis and inhibiting adipocyte formation through involving Wnt5a and NFATc signaling pathway.

The main focus of the paper is to propose a smart recommender system based on the methods of hybrid learning for personal well-being services, called a smart recommender system of hybrid learning (SRHL). The essential health factor is considered to be personal lifestyle, with the help of a critical examination of various disciplines. Integrating the recommender system effectively contributes to the prevention of disease, and it also leads to a reduction in treatment cost, which contributes to an improvement in the quality of life. At the same time, there exist various challenges within the recommender system, mainly cold start and scalability. To effectively address the inefficiencies, we propose combined hybrid methods in regard to machine learning. The primary aim of this learning method is to integrate the most effective and efficient learning algorithms to examine how combined hybrid filtering can help to improve the cold star problem efficiently in the provision of personalized well-being in regard to health food service. These methods include: (1) switching among content-based and collaborative filtering; (2) identifying the user context with the integration of dynamic filtering; and (3) learning the profiles with the help of processing and screening of reflecting feedback loops. The experimental results were evaluated by using three absolute error measures, providing comparable results with other studies relative to machine learning domains. The effects of using the hybrid learning method are gathered with the help of the experimental results. Our experiments also show that the hybrid method improves accuracy by 14.61% of the average error predicted in the recommender systems in comparison to the collaborative methods, which mainly focus on the computation of similar entities.