• KIPS Transactions on Computer and Communication Systems
• /
• v.11 no.12
• /
• pp.437-444
• /
• 2022

Recently, research on mobile edge services has been conducted to handle computationally intensive and latency-sensitive tasks occurring in wireless networks. However, MEC, which is fixed on the ground, cannot flexibly cope with situations where task processing requests increase sharply, such as commuting time. To solve this problem, a technology that provides edge services using UAVs (Unmanned Aerial Vehicles) has emerged. Unlike ground MEC servers, UAVs have limited battery capacity, so it is necessary to optimize energy efficiency through load balancing between UAV MEC servers. Therefore, in this paper, we propose a load balancing technique with consideration of the energy state of UAVs and the mobility of vehicles. The proposed technique is composed of task offloading scheme using genetic algorithm and task migration scheme using Q-learning. To evaluate the performance of the proposed technique, experiments were conducted with varying mobility speed and number of vehicles, and performance was analyzed in terms of load variance, energy consumption, communication overhead, and delay constraint satisfaction rate.

• Journal of Internet Computing and Services
• /
• v.21 no.5
• /
• pp.1-7
• /
• 2020

Deep neural networks(DNN), which are used as approximation functions in reinforcement learning (RN), theoretically can be attributed to realistic results. In empirical benchmark works, time difference learning (TD) shows better results than Monte-Carlo learning (MC). However, among some previous works show that MC is better than TD when the reward is very rare or delayed. Also, another recent research shows when the information observed by the agent from the environment is partial on complex control works, it indicates that the MC prediction is superior to the TD-based methods. Most of these environments can be regarded as 5-step Q-learning or 20-step Q-learning, where the experiment continues without long roll-outs for alleviating reduce performance degradation. In other words, for networks with a noise, a representative network that is regardless of the controlled roll-outs, it is better to learn MC, which is robust to noisy rewards than TD, or almost identical to MC. These studies provide a break with that TD is better than MC. These recent research results show that the way combining MC and TD is better than the theoretical one. Therefore, in this study, based on the results shown in previous studies, we attempt to exploit a random balance with a mixture of TD and MC in RL without any complicated formulas by rewards used in those studies do. Compared to the DQN using the MC and TD random mixture and the well-known DQN using only the TD-based learning, we demonstrate that a well-performed TD learning are also granted special favor of the mixture of TD and MC through an experiments in OpenAI Gym.

• KNOM Review
• /
• v.22 no.3
• /
• pp.13-19
• /
• 2019

Network anomaly detection is a technology that collects information about flows on a network and detects malicious attacks occurring in a network in real time. In-band Network Telemetry (INT) technology provides more detailed information in real time, that is not provided by existing networks, such as hop latency and queue occupancy. In this paper, we propose the method to implement an anomaly detection system with higher performance by using INT as an input feature of machine learning and verify it through experiments.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.34 no.6
• /
• pp.403-408
• /
• 2021

Currently, in the design process of civil structures such as bridges, it is common to make final products by repeating the process of redesigning, if the initial design is found to not meet the standards after a structural review. This iterative process extends the design time, and causes inefficient consumption of engineering manpower, which should be put into higher-level design, on simple repetitive mechanical work. This problem can be resolved by automating the design process, but the external analysis program used in the design process has been the biggest obstacle to such automation. In this study, we constructed an AI-based automation system for the bridge design process, including an interface that could control both a reinforcement learning algorithm, and an external analysis program, to replace the repetitive tasks in the current design process. The prototype of the system built in this study was developed for a 2-span RC Rahmen bridge, which is one of the simplest bridge systems. In the future, it is expected that the developed interface system can be utilized as a basic technology for linking the latest AI with other types of bridge designs.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.28 no.6
• /
• pp.1463-1474
• /
• 2018

Fuzzing is one of the software testing techniques that find security flaws by inputting invalid values or arbitrary values into the program and various methods have been suggested to increase the efficiency of such fuzzing. In this paper, focusing on the existence of field with high relevance to coverage and software crash, we propose a new method for intensively fuzzing corresponding field part while performing field based fuzzing. In this case, we use a deep learning model called Variational Autoencoder(VAE) to learn the statistical characteristic of input values measured in high coverage and it showed that the coverage of the regenerated files are uniformly higher than that of simple variation. It also showed that new crash could be found by learning the statistical characteristic of the files in which the crash occurred and applying the dropout during the regeneration. Experimental results showed that the coverage is about 10% higher than the files in the queue of the AFL fuzzing tool and in the Hwpviewer binary, we found two new crashes using two crashes that found at the initial fuzzing phase.