Earticle

Recent ubiquitous computing environments increasingly impact on our lives using the current technologies of sensor network and ubiquitous services. In this paper, we propose ubiquitous architectural framework for ubiquitous sleep aid service(UbiSAS) in the subset of ubiquitous computing for refreshing of human's sleep. And we examine technical feasibility. Human can recover his health through refreshing sleep from fatigue. Ubiquitous architectural framework for UbiSAS in digital home offers agreeable sleeping environment and improves recovery from fatigue. So we present new concept of ubiquitous architectural framework dissolving stress. Specially, we apply context to context-aware framework module. This context is transferred to context adaptive inference engine which has service invocation function in intelligent agent module. Ubiquitous architectural framework for UbiSAS using context adaptive rule inference engine without user intervention is technical issue. That is to say, we should take sleep comfortably during our sleeping. And sensed information during sleeping is changed to context-aware information. This presents significant information in context adaptive rule inference engine for UbiSAS. This information includes all sleeping state during sleeping in context-aware computing technique. So we propose more effective and most suitable ubiquitous architectural framework using context adaptive rule inference engine for refreshing sleep in this paper.

현재 스마트폰의 급격한 보급과 IoT을 대상으로 활성화로 인해 소셜네트워크 서비스를 이용하여 악성코드 를 유포하거나 지능화된 APT와 랜섬웨어 등과 같은 지능적인 침입이 진행되고 있고 이로 인한 피해도 이전의 침입 보다는 많이 심각해지고 커지고 있는 실정이다. 따라서 본 논문에서는 이런 지능적인 악성 코드로 이루어지는 침입 행위를 탐지하기 위하여 지능적인 침입 상황 인식 추론 시스템을 제안하고, 제안한 시스템을 이용하여 지능적으로 진행되는 다양한 침입 행위를 조기에 탐지하고 대응하게 하였다. 제안 시스템은 이벤트 모니터와 이벤트 관리기, 상황 관리기, 대응 관리기, 데이터베이스로 구성되어 있으며 각 구성 요소들 사이에 긴밀한 상호 작용을 통해 기존 에 인식하고 있는 침입 행위를 탐지하게 하고 새로운 침입 행위에 대해서는 학습을 통해 추론 엔진의 성능을 개선 하는 기능을 통하여 탐지하게 하였다. 또한, 지능적인 침입 유형인 랜섬웨어를 탐지하는 시나리오 통하여 제안 시 스템이 지능적인 침입을 탐지하고 대응함을 알 수 있었다.

At present, due to the rapid spread of smartphones and activation of IoT, malicious codes are disseminated using SNS, or intelligent intrusions such as intelligent APT and ransomware are in progress. The damage caused by the intelligent intrusion is also becoming more consequential, threatening, and emergent than the previous intrusion. Therefore, in this paper, we propose an intelligent intrusion situation-aware reasoning system to detect transgression behavior made by such intelligent malicious code. The proposed system was used to detect and respond to various intelligent intrusions at an early stage. The anticipated system is composed of an event monitor, event manager, situation manager, response manager, and database, and through close interaction between each component, it identifies the previously recognized intrusive behavior and learns about the new invasive activities. It was detected through the function to improve the performance of the inference device. In addition, it was found that the proposed system detects and responds to intelligent intrusions through the state of detecting ransomware, which is an intelligent intrusion type.

The purpose of this paper is to develop effective learning strategies to improve lexical ability as well as reading ability. The strategies investigated in this paper and the conclusions drawn from this study will provide EFL learners and teachers with a better understanding of how effective lexical strategies can lead to improvements in language learning. To this end this paper will focus on 4 areas of research as a means to determine the importance of context in the application of various lexical inference strategies. In this study we will show (i) some effective ways to increase the success ratio of lexical inference strategies and (ii) some teaching/learning methods to improve lexical ability.

ChatGPT의 등장 이후 대규모 언어 모델(Large Language Model, LLM)을 다양한 분야에 적용하는 연구가 활발하게 이루어지고 있다. LLM의 추론 성능을 향상시키기 위한 대표적인 방법으로는 Chain-of-Thought, Retrieval-Augmented Generation(RAG) 등이 있다. 그러나 이러한 방법들은 입력 프롬프트(Prompt)의 길이를 증가 시켜 추론 시간과 비용을 높이며, 불필요한 문맥으로 인해 환각 현상이 발생하는 문제가 있다. 본 논문에서는 이러한 문제를 해결하기 위해, 프롬프트 중 질의문을 기준으로 관련된 정보만 남기고 문맥(Context)을 압축하는 문맥 필터링 (Context Filtering) 기법을 제안한다. 문맥 필터링은 질의문의 시계열적, 의미적 관련성을 기준으로 필요한 정보만 입력 으로 사용하여 LLM이 추론하는 방법이다. 실험은 전장 시뮬레이션으로부터 생성된 Timestamp + Triplet 구조의 대규 모 시계열 데이터를 기반으로 수행되었다. 실험 결과, 전체 문맥의 약 25%만을 사용한 압축 프롬프트에서도 기존 프롬프 트와 유사한 수준의 상황 인지 정확도를 유지하였으며, 추론 시간과 토큰 사용량 또한 감소하였음을 확인하였다.

Since the advent of ChatGPT, research on applying large language models(LLM) to various fields has been actively conducted. Typical methods for improving LLM's reasoning performance include Chain-of-Thought and Retrieval-Augmented Generation(RAG). However, these methods increase the length of the input prompt, thereby increasing the inference time and cost, and there is a problem that Hallucination occurs due to unnecessary context. In order to solve this problem, this paper proposes a Context Filtering technique that compresses context, leaving only relevant information based on the questionnaire among prompts. Context Filtering is a method for LLM to infer using only necessary information as input based on the time-series and semantic relevance of a query. The experiment was conducted based on time-series data of Timestamp + Triplet structure in a battlefield situational awareness scenario. As a result of the experiment, it was confirmed that the situation awareness accuracy similar to that of the existing prompt was maintained even at the compression prompt using only about 25% of the entire context, and the inference time and token usage were also reduced.

In the IOT environment, sensor data stream consists of event data from heterogeneous multi-sensors. One type of sensor may have quite a different event frequency from those other kinds of sensors, which makes most sensor data sets imbalanced. To classify an imbalanced data effectively, it is necessary to preprocess it for converting into a balanced data. This process may unify heterogeneous attributes in the imbalanced data and alleviate the difficulties for data mining on it. Mass function plays an important role in the fuzzy theory and Dempster-Shafer Theory. In this paper, using a mass function is suggested to process imbalanced data stream. A mass function is developed to compute mass values for imbalanced data sets, and an experiment is performed to investigate the validity to apply the mass function to the sensor data stream.

The Basic Probability Assignment is also called mass function. The BPA should basically include the uncertainty and obscurity of the context of the real world to recognize and reduce the uncertainty in the result of the arithmetic operation. The BPA makes a decision to make it possible the arithmetic operation between different heterogeneous signals or data. This study proposed calculation of the BPA by getting a clue from a policy that recognizes at a certain time interval and reports changes worth reporting. Usually, ‘the size of changes’ is often the base in a decision of meaningful recognition of events reported to the host. Like this, what aspect the value sensed with the lapse of time has as compared to that sensed in previous time is related to the context. Thus, in deciding the BPA, reflecting this point is relevant. This study analyzed the change of the proposed sensing value to decide the BPA, and based on this, it proved that contexts could be inferred.

Multi-sensors data fusion takes center stage in wireless sensor network system for acquiring more precise information. The event information detected and reported by sensors is often not reliable to infer contexts. The study presents that the data are evaluated before the data fusion on all event information acquired is occurred in the aggregate. The evaluation of event information is to estimate if context inference is possible and to reduce the processes of unnecessary multi-sensors data fusion. The study shows the frequency of event occurrences and the average variation rates of the data reported by the sensors for evaluating event information as the evaluation criteria. As a result of evaluating event information based on the suggested evaluation criteria and processing data fusion selectively, unnecessary data fusion processes were reduced, the data processes for context inference were decreased, and eventually the efficiency of context inference was improved.

The things to realize using the network system of ubiquitous sensors are situation perception and personalization services. To enhance performance of the situation perception, construct wireless networks with multi-sensors. In other to enhance the performance of the context awareness, temporary storage and utilization for the sensing data were proposed. The criteria to judge whether to use or not the sensing data reported from the sensor in the next time zones was presented by saving it temporarily, the data proved to be used in the next time zones should be recycled. On using by calling the Basic Probability Assignment saved, weights corresponding to the increment of the previous and present time zones should be added. Since then, the belief and uncertainty were calculated by using each focal element's BPA and through this, advanced information can be deduced rather than obtaining context information.

We propose architecture for multi-agent systems to retrieve and classify features extracted from images and videos of smart cameras. To enable cooperative inference between agents on cameras, structured representation of agents’ knowledge and abilities is required in the form of ontologies. Recognized features if properly structured and annotated, can be a useful source of information for context aware surveillance. This work builds a hierarchical inference data deployment structure and import related and required data to annotate rich data arriving from multiple sensor streams, in this case smart cameras. The annotation provides an impetus to the improvement of knowledge over time. Proactive deployment provides the main concepts and properties to model a hierarchical area ontology structure which can span a university campus or a city. We also define management policies to compare their performance for the wide area surveillance specifically.

It is not enough to recognize the situations which currently occur simply. The current situations have the causes that they get to occur. The causes can just be generated and they can make the situations like the current states while the ones which occurred in the past have continued. Furthermore, if the causes which made the current situations don't disappear, they can continue to stay the same, get worse, or be changed to another situation. Therefore, limiting the range of context awareness to the situations which currently occur can be insufficient as the system which recognizes situations of the everyday world. Therefore, this study aims at problem-solving of two things. First, it recognizes situations without advance information. Second, it infers causes of situations and predicts how the situations will turn out in the future. To solve these problems, this study uses multiple sensor data fusion together using Dempster-Shafer Evidence Theory (DST) and Kalman Filter (KF). It recognizes situations under the conditions without any advance information through DST, infers causes of the current situations, and predicts how the current situations will turn out in the future. At this moment, BPA is important to recognize situations through DST and infer causes and state transition equation plays an important role in predicting arrangement through KF. The study carries out context inference and cause inference using DST. It describes the plan which infers causes of situations without advance information. It calculates required state transition equation to predict the progress of the research and infers how the causes revealed through DST by using it will arrange the current situations in the future by using KF.

Measures for inferring context through multiple sensor data fusion are being actively used in various fields. The landslide protection wall refers to the structure/engineering method that prevents the collapse of soil during ground excavation. To prevent safety accidents of the landslide protection wall, it is necessary to devise measures for detecting the deformation of the landslide protection wall and for inferring the cause of the deformation. Context inference through multiple sensor data fusion can be used to analyze the influential factors of the problem among safety-threatening factors during generation of anomalous events in the landslide protection wall. This study presents the method for inferring decisive factors through multiple sensor data fusion based on the Dempster-Shafer evidence theory in an environment in which the deformation of the landslide protection wall is affected by earth pressure, water pressure and surcharge effect. Based on the Basic Probability Assignment (BPA) function of factors, this study calculated the belief and plausibility of factors through Dempster’s Rule of Combination. Uncertainty intervals can be calculated and compared with one another based on the belief and plausibility of factors to achieve inference of decisive factors

In an environment in which several events are sensed in a complex manner and sequentially obtained, a clue can be obtained for inference of situations by classifying each event and analyzing the aspect of change of each event. The study proposes a method to efficiently decide the cluster centers in each subsequent time slot for efficient classification of events and inference of situations in a data stream environment. For the data stream under this condition, each time slot classified at a certain interval is set up, the events using clustering in each time slot are carried out, and to recognize how the aspect of change of each event sensed in a continuous time slot is carried out, the cluster centers are allowed to be rapidly captured.

Big Data in the form of streaming, the data stream mining, has received a great deal of attention for some time. With disparate multiple sensors, sensors are able to gentrify the information they want to acquire. In this paper, the ways to encode a wide range of sensor data that is continuously reported is proposed. These encoding methods enable higher level analysis than identify the frequent pattern or association rules. It is essential that sensors are distributed and extract various and detailed information about the context that was sensed. This study suggests that it is essential that the sensor data is encoded in a reasonable and valid way of context inference and extracting a variety of quality information even in the data stream environment for on-off analysis of the large amount of sensor data that continuously flow in through the sensor data encoding method.

Dempster-Shafer Evidence Theory (DST) plays an important role in multi-sensor data fusion. In this paper, we propose weighting method based on event information sensor of aggregation for context awareness in wireless sensor networks. Context inference in wireless sensor networks has to use weighting method, and context inference based on sensor aggregation due to structure feature of wireless sensor network. We propose weight method based on event frequency of the sensor and the repetitive of event report sensor at the continuous time slot. The validity of reasoning through context inference in wireless sensor network can improve.

Dempster-Shafer Evidence Theory can infer the context only with some native sensed signals but its number of assumptions is limited, compared to the number of evidence. In this research, we proposed a method that can increase the number of sensors collecting symptoms for more dynamic contextual inference yet with reduced calculation load increase. This is possible by clustering sensors with relevance to a situation and fusing data hierarchically.

Basic probability assignment (BPA) has the key role in multi-sensor data fusion. This paper proposed a way to determine BPA using the various signals acquired from the sensors. It described the analysis of signals detected by the sensors. The determined BPA were used for multi-sensor data fusion to infer and recognize the context targeted by a wireless sensor networks. To determine BPA, the change rate was calculated and assessed to be reflected. The method enabled context inference using the sensor signals even when there was no advanced information of the situation.

Smart home services for users require precise and reliable context information to guarantee credibility. To do this, system should have ability of context-awareness. Most of the context-aware systems have adopted rule-based inference but it did not fulfill users’ demand. To support valuable services in smart home environment, we present the rule-based inference engine which uses user profiles to let the system be aware of users’ context much accurately. Since the physical and logical size of the domain is relatively small, we also introduce inference algorithm optimized for limited size of memory space of the system.

넓은 지역을 관찰하는 상황 인지 시스템은 보다 효율적이고 정확한 상황 인지를 위하여 각 로컬 시스템의 추론시스템을 연결하는 협동적 상황 인지 시스템을 구성한다. 본 논문에서는 물리보안 감시 시스템을 위한 온톨로지 기반 에이전트의 설계 및 구현 내용을 보인다. 다양한 영상 및 이벤트 소스 가 넓은 지역에 분포하는 경우, 중앙 집중화된 거대시스템을 구성하는 것이 비효율적이거나 불가능하 다. 각 데이터 소스가 독립적인 에이전트를 두어 로컬 서비스와 더불어 동등 수준 혹은 상하위 시스 템과의 통신을 통한 각 서비스의 융합 및 2차 추론을 거친 고수준 서비스를 제공하여야 한다. 이를 위해서 온톨로지를 기반으로 지능적 추론을 제공하는 에이전트를 설계 및 구축하였다. 또한 보다 효 율적인 추론 시스템을 위하여 추론에 쓰이는 온톨로지를 동적인 중요도 평가에 따라 재구성하여 추 론에 적용함으로써 상황인지 및 대응에 보다 효율적으로 이용되도록 하였다.

For more efficient and accurate context awareness, wide area context awareness system constructed as a cooperative context-aware system. This paper shows the design and implementation of the agents based on ontology for physical security. When various video and event sources are distributed wide, it is inefficient or impossible to build a centralized mega systems. Each data source should have independent agent and provide high level service based on knowledge fusion and inference between peer systems or other neighbor systems. To accomplish our goal, we have designed and implemented agents which provide intelligent inference based on ontology. Furthermore for more efficient inference systems, ontologies are restructured reflecting dynamic weight change and applied to it resulting in better context awareness and response.

Currently, web applications are gaining in prevalence. In a web application, an input may not be appropriately validated, making the web application susceptible to cross-site scripting (XSS), which poses serious security problems for Internet users and websites to whom such trusted web pages belong. A taint inference is a type of information flow analysis technique that is useful in detecting XSS on the client side. However, in existing techniques, two current practical issues have yet to be handled properly. One is URL rewriting, which transforms a standard URL into a clearer and more manageable form. Another is HTML sanitization, which filters an input against blacklists or whitelists of HTML tags or attributes. In this paper, we make an analogy between the taint inference problem and the molecule sequence alignment problem in bioinformatics, and transfer two techniques related to the latter over to the former to solve the aforementioned yet-to-be-handled-properly practical issues. In particular, in our method, URL rewriting is addressed using local sequence alignment and HTML sanitization is modeled by introducing a removal gap penalty. Empirical results demonstrate the effectiveness and efficiency of our method.