Vol. 47, No. 5,
May 2020
Digital Library
Visualization of Convolutional Neural Networks for Time Series Input Data
http://doi.org/10.5626/JOK.2020.47.5.445
Globally, the use of artificial intelligence (AI) applications has increased in a variety of industries from manufacturing, to health care to the financial sector. As a result, there is a growing interest in explainable artificial intelligence (XAI), which can provide explanations of what happens inside AI. Unlike previous work using image data, we visualize hidden nodes for a time series. To interpret which patterns of a node make more effective model decisions, we propose a method of arranging nodes in a hidden layer. The hidden nodes sorted by weight matrix values show which patterns significantly affected the classification. Visualizing hidden nodes explains a process inside the deep learning model, as well as enables the users to improve their understanding of time series data.
The Classification Model of Fileless Cyber Attacks
GyungMin Lee, ShinWoo Shim, ByoungMo Cho, TaeKyu Kim, KyoungGon Kim
http://doi.org/10.5626/JOK.2020.47.5.454
Since late 2000, state-sponsored and sophisticated cyber-attacks have continued unabated. Also, preparing countermeasures against cyber-attack techniques based on traces are also in development. Cyber attackers use a variety of techniques to veil their attacks from these analyses and countermeasures. In particular, fileless cyber-attacks that do not create a file used for an attack are increasing. Fileless cyber-attacks are difficult to analyze because there are no executable files to analyze from the defender"s perspective. In this paper, we investigate and analyze fileless cyber-attacks and present a model based on the cyber kill chain to classify fileless cyber-attacks. Through this, it is expected to identify and respond to attack types more quickly than when new fileless cyber-attacks occur.
Optimizing Swap Use of Programs Using Memory Access Profiling
Yunjae Lee, Heon Y. Yeom, Hyuck Han
http://doi.org/10.5626/JOK.2020.47.5.466
The slow growth rate of main memory and modern computing workloads requires lots of memory, making main memory the bottleneck of system performance. Swapping provides a large virtual memory to programs by utilizing fast but small main memory and large secondary storage. However, programs cannot accomplish optimal performance due to conservative swapping policy which targets general workloads. The objective of this study was to analyze memory access pattern of programs and optimize programs to utilize swapping considering memory access pattern. A low-overhead memory profiling technique and a simple optimization technique can help programmers optimize their programs with ease. We optimized six workloads using these techniques and improved the performance of the workloads by 43%.
A Knowledge Graph Embedding-based Ensemble Model for Link Prediction
http://doi.org/10.5626/JOK.2020.47.5.473
Knowledge bases often suffer from their limited applicability due to missing information in their entities and relations. Link prediction has been investigated to complete the missing information and makes a knowledge base more useful. The existing studies on link prediction often rely on knowledge graph embedding and have shown trade-off in their performance. In this paper, we propose an ensemble model for knowledge graph embedding to improve quality of link prediction. The proposed model combines multiple knowledge graph embeddings that have unique characteristics. In this way, the ensemble model is able to consider various aspects of the entries within a knowledge base and reduce the variation of accuracy depending on hyper-parameters. Our experiment shows that the proposed model outperforms other knowledge graph embedding methods by 13.5% on WN18 and FB15K dataset.
Analysis of Speech Emotion Database and Development of Speech Emotion Recognition System using Attention Mechanism Integrating Frame- and Utterance-level Features
http://doi.org/10.5626/JOK.2020.47.5.479
In this study, we propose a model consist of BLSTM (Bidirectional Long-Sort Term Memory) layer, Attention mechanism layer, and Deep neural network to integrate frame- and utterance-level features from speech signals model reliability analysis the labels in the speech emotional database IEMOCAP (Interactive Emotional Dyadic Motion Capture). Based on the evaluation script of the labels provided in the IEMOCAP database, a default data set, a data set with a balanced distribution of emotion classes, and a data set with improved reliability based on three or more judgments were constructed and used for performance of the proposed model using speaker independent cross validation approach. Experiment on the improved and balanced dataset achieve a maximum score of 67.23% (WA, Weighted Accuracy) and 56.70% (UA, Unweighted Accuracy) that represents an improvement of 6.47% (WA), 4.41% (UA) over the baseline dataset.
Rules-based Korean Dependency Parsing Using Sentence Pattern Information
Sung-Tae Kim, Minho Kim, Hyuna Kim, Hyuk-Chul Kwon
http://doi.org/10.5626/JOK.2020.47.5.488
The parser proposed in this paper is a wide range dependency parser that facilitates dependency r-elations to all the possible candidates appearing in sentences. Output a parse tree of all candidates appearing in a sentence in which neutrality can occur, and use the rules to advance the ranking. Use the agenda mechanism to form a dominance-dependency relationship with the graph analysis method and create a candidate tree from the input sentence through the four stages of the analysis process. Additionally, for the proper use of sentence pattern information corpus, we implemented rules and algorithms that overcome the limitations of previous studies and enhanced the ranking of candidate parse trees using the sentence pattern information. As well as difficulty in ranking the [noun - determiner] strengthened the ranking using sentence pattern information about qualities. As a result, the UAS (unlabeled attachment score) of the parse tree top-rank improved by 0.74%p, and the average correct ranking of the candidate tree improved by 28.1%. Additionally, the highest performance was UAS 94.02%.
Mechanized Proof of Type Preservation for Polymorphic Lambda Calculi Using Abella
http://doi.org/10.5626/JOK.2020.47.5.496
Abella is a proof assistant with several attractive merits including the support for unification over lambda-tree syntax or higher-order abstract syntax (HOAS); due to such merits, a concise solution for the POPLmark Challenge 2A, i.e., mechanized proof of the type preservation for the pure fragment of the call-by-value System F<:, has been developed in Abella. In this work, we explain our completed mechanized proof for the type preservation of the polymorphic lambda-calculus (System F), not necessarily limited to a certain evaluation strategy, and report the status of our ongoing work extending the type preservation proof for the higher-order polymorphic lambda-calculus (System Fω).
Detection of Power Contract Violations using an Anomaly Pattern Detection Method on Power Consumption Data Streams
http://doi.org/10.5626/JOK.2020.47.5.504
Due to the charging system for electricity in which the amount of money charged varies depending on the purpose it is used for, as laid out by the contract between the electricity provider and the consumer, any losses caused by contract violation can be greatly increased. Recently, with the expansion of the Advanced Metering Infrastructure (AMI), it has become possible to collect the electricity usage information of consumers in real time through smart meters. However, studies into the detection of violations, including contract type violations, still have difficulties caused by data imbalances due to a lack of collected violation data. In this paper, we propose a method to detect contract type violations by modelling normal usage patterns based on normal consumers’ usage data according to specific contract type: from this base we apply an abnormal pattern detection method to the AMI smart meter data stream of consumers with a particular contract type. In an experiment simulating contract violations using power usage data of about 300 people collected from smart meters over a 3 years 7 months period, the proposed method obtained an f1 value of 0.83 and a detection delay of, on average, 6 days after violation occurred. This shows that the proposed method can be effectively used for detecting contract violation in real situations.
Single Group Collective Trip Planning Query Processing Using G-tree Index Structures on Road Networks
http://doi.org/10.5626/JOK.2020.47.5.513
In this paper, we discuss Single Group Collective Trip Planning (SGCTP) queries that minimize the overall travel cost in location-based ride sharing services. The SGCTP queries identify a meeting point that minimizes the overall cost of such a trip when a group of users are gathered at a particular point and travel to the destination using one vehicle. Although many researches on collective trip planning queries have been conducted, there is a problem that the query performance is effective only in a specific situation. So, we introduce a baseline method of the SGCTP queries and then, propose an effective pruning technique with a G-tree index structure. Additionally, we analyze that the limitations of the previous studies, and experimental results show that the proposed pruning technique can obtain the optimal query result without being affected by the limitations of the previous studies.
Automatic Data Plane Slice Provisioning on Mobile-CORD
Javier Diaz Rivera, Wang-Cheol Song
http://doi.org/10.5626/JOK.2020.47.5.526
The goal of the 5G mobile network is to provide the right ecosystem for diverse and heterogeneous user equipment (UE) connections. As this ecosystem contains multipurpose UE, the data plane that serves the transmission of information must be tailored to accommodate the services that are being consumed by users. By using Network Slicing, Software-Defined Networking (SDN), Network Functions Virtualisation (NFV) and a platform for Service Orchestration, this paper showcases a design for a Data Plane Provisioning Module that can adjust the Data Plane Slices of an Evolved Packet Core (EPC) system on the basis of QoS policies defined as high-level contracts on the Application Layer of a Three-Tiered Architecture.
SDN-based Task Allocation for IoT-Fog Network
Dzaky Zakiyal Fawwaz, Sang-Hwa Chung
http://doi.org/10.5626/JOK.2020.47.5.535
The Internet of Things requires that resources are allocated for it to execute tasks. Utilizing fog computing may have many benefits in these circumstances since it offers distributed resources that could give lower latency, lower bandwidth, and various other advantages compared to cloud computing. If we wish to use fog computing, we need to consider how to perform task allocation over multiple fog nodes. The kind of networks used are likely to have not only continuous incoming IoT tasks but also other dynamic network conditions. Hence, we introduce dynamic task allocation that utilizes a Software-Defined Network. Our system handles each incoming task by considering network and fog node statistics. The task allocation method must select the optimal pair of fog nodes and also the path because there are multiple fog nodes and many feasible related paths to deal with. Thus, we define the problem to be one of finding the multi-source, single-target, shortest path on a network graph, to help solve the problem we formulate the joint fog node-link utilization cost. We also propose a Many-to-One Shortest Path algorithm to solve such a problem. The experiments we performed to evaluate our system show that it outperforms the previous state-of-the-art work. Averaging over all the experiment"s topologies, our method achieves higher tasks per second completion rate, a lower response time and lower fog node/link utilization with scores of 37 tasks/sec, 676ms and 65%/24% utilization, respectively.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr