ゼミナール発表

日時: 5月26日(月)3限 (13:30-15:00)


会場: L1

司会: 畑 秀明
岡本 薫 1351201: M, 1回目発表

title: Investigate Cyber-attacks using a Multicopter

abstract: In this presentation, investigate the ability of a system to detect cyber-attacks using a Multicopter. The recent advent and popularity of the Multicopter has also raised the possibility of new type of the cyber-attacks which can be carried out by such a device. Until now, cyber-attacks were generally carried out via the internet. However, new types of the cyber-attacks are carried out using wireless networks from nearby systems. We selected the AR. Drone 2.0 as a research platform, and we conducted a survey analysis. Based on the results, we can use the complex information of network traffic, radio waves, image and sound in order to create an “Integrated Surveillance System for the Multicopter.“ As an example of a detection technique, by analyzing the packets that originated from AR. Drone 2.0 obtained by the system, we were able to show that the detection of access to the system and to recognize whether the manual operation or automatic pilot.*

language of the presentation: Japanese

発表題目: マルチコプターを用いたサイバー攻撃に対する一検討

発表概要: 本発表は,マルチコプターを用いたサイバー攻撃の検知方法について検討を行う.小型コンピュータを搭載したマルチコプターの登場により,これまでのサイバー攻撃のようにネットワーク経由ではなく,物理的に攻撃対象に接近し,建物外部から無線ネットワークを介してサイバー攻撃を行うことが可能となりつつある.本研究では,マルチコプターとしてAR. Drone 2.0を対象にサイバー攻撃の実行可能性の調査・分析を行った結果をもとに,対象ネットワークの有線・無線ネットワークトラフィック,電波,映像,音の複合情報を用いたマルチコプター用統合監視システムについて検討を行う.また,検知手法の一例として,無線ネットワーク上でのAR. Drone 2.0の通信パケットを取得,分析することで,AR. Drone 2.0の接近及び手動操縦と自動操縦の判別が可能であることを示した.

 
JIN YONG 1351207: M, 1回目発表
title: Finding Origin of Patches Included by Linux Distributions
abstract: A Linux distribution is a Linux system that can be easily used after installing. There are lots of Linux distributions, like Ubuntu, Redhat, CentOS and so on nowadays, and they apply patches by their criteria. For example, when Radhat Enterprise Linux (RHEL) make packages for one software, first, they pick one tag version from the softwareソs repository as base version, then, pick modifications they want to apply from the repository, finally, add modifications made by them to make patches, and apply them to packages. So, when developer wants to make there own modifications to the software, they have no idea with what kind of modifications were made. What I try to do is find origin of patches in order to provide detailed information to developers, so the developers can do what they want without uncertainty. I will talk about the process of finding origin of patches, and the results of this timeソs investigation, and about future work.
language of the presentation: Japanese
 
KAR LONG CHAN 1351203: M, 1回目発表
title: (Paper Introduction)"GamingAnywhere: An Open Cloud Gaming System"
In gaming industry, cloud gaming is a new cloud service on trend trying to serve millions of players around the world with novel gaming experience. It aims to provide high-quality gaming service at any device, including thin clients, which are incapable of handling high-definition game softwares. Players are only required to use any device that can display game contents and be able to connect to cloud servers for receiving data through network. Ideally it is a promising gaming service but in reality, various technologies make cloud gaming not comprehensively feasible for every type of game, such as first person shooting game which requires fast responsiveness. QoE will decrease dramatically if response from the cloud system is not rapid enough. In addition, existing cloud service, which mostly streams encoded video sequence back to the client, is hard to catch up with the rising demands for Graphic Quality. Furthermore, most existing cloud gaming platforms are close source with proprietary control, which are not open for development. In this presentation, GamingAnyWhere, which is the first and the only open source cloud gaming platform will be introduced. In addition, a hybrid cloud gaming platform will be proposed for increasing efficiency and graphic quality for cloud gaming.
language of the presentation: English
 
PONGSAKORN U-CHUPALA 1351211: M, 1回目発表
Title: Overseer: Bandwidth and Latency Aware Routing using OpenFlow
Abstract: Each application running on a network infrastructure may have different requirements. Some performs better at low latency while the others benefits from having higher bandwidth. It is possible to dynamically control routes (flows) in the network using Software-Defined Networking (SDN) technology. With these 2 facts, we aim to improve network utilization by introducing bandwidth and latency aware routing mechanism. We proposed SDN controller design which take information about current status of the network such as bandwidth and latency of each link in real time from monitoring tool and allow application to define flow preference for each connection. The controller then uses this information to determine route for each connection accordingly.
language of the presentation: English
 
ZHANG ANNA 1351213: M, 1回目発表
title: Accelerator on computing architecture to accelerate algorithms with high performance and low energy
abstract: Introduction of one "Best Paper" about how an efficient accelerator working on computing architecture, which can accelerate advanced Artificial Neural Network algorithms with high performance, low engrgy and small space.
language of the presentation: English
 

会場: L2

司会: Kevin Duh
DO QUOC TRUONG 1351206: M, 1回目発表
title: Paralinguistic Speech Translation
Speech-to-speech (s2s) translation is gradually starting to break down the language barrier, bringing the the opportunities for people to understand each other using different languages. However one of the limitations of current S2S systems that they usually do not translate the paralinguistic information included in the input speech. Among the various types of paralinguistic information, we focus on emphasis, a type of information that is used to convey the focus of the sentence, emotion of the speaker, or other high level information useful for communication. In this presentation, I will talk about a method to translate paralinguistic speech.
language of the presentation: English
 
LIANG JUN 1351208: M, 1回目発表
title: Phrase-based Bilingual Lexicon Extraction with Topic Model
abstract: The traditional lexicon-based miachine translation is based on word-to-word translation. To a certain degree, it is suitable for translating phrases. Meawhile, it causes polysemous problem. To solve this kind of problem, the proposed approach focuses on extracting a phrase-to-phrase bilingual lexicon to be more suitable for translating full sentences.
language of the presentation: English
 
LUIZ GUSTAVO MOREIRA SAMPAIO 1351209: M, 1回目発表
title: Multi display management with mobile devices
abstract: We propose a window management system for multi display environment, using embedded patterns that are invisible to human eye, but easily recognized by mobile device cameras. With this approach we can control the elements on different displays, without affecting the visibility of the content presented.
language of the presentation: English
 
PHILIP ARTHUR 1351210: M, 1回目発表
title: The extraction of 3-Synchronous Grammar - Keyword, Paraphrase, and Logical Form
Synchronous Context Free Grammar, is high level abstraction of parsed structure of the sentence which is widely used in many Natural Language Processing' application. This abstraction is used to recover an underspecified input' structure and its cannonical form. To ensure fluent output from parsing, language model is used to ammeliorate the error of the parser by taking the best probability of the paraphrase output context into account.
language of the presentation: English
 

会場: L3

司会: 小野 直亮
ABHINAV DADHICH 1351202: M, 1回目発表
title: Dynamic Object extraction in 3D occupancy grid using Hidden Markov Model
abstract: For a Mobile Service Robots, it is necessary for the robot to navigate in a dynamic environment and also interact with people. To effectively interact with people, robot needs to classify human from other objects. In this Seminar, method of categorizing dynamic objects from the static parts in environment is proposed.Occupancy grid is a robust method to map real world data from sensors for probabilistic update.Proposed method uses a 3D occupancy grid with Octree structure to update the occupancy in real-time. While, Hidden Markov Model is used to categorize cells in occupancy grid which are occupied by dynamic object.
language of the presentation: English
 
CHRISTIAN DEUS TELMO CAYAO 1351204: M, 1回目発表
title: Stereo Vision System with Adaptive Air-Water Interface Refraction Correction for Robot Manipulation
abstract: Vision is an effective sensing form for humans and animals in understanding their environment, recognition of the objects around them and interacting with the environment. Stereo vision serves as a sensing unit for robots which can perceive depth using a pair of eyes like humans and animals. This seminar considers robot vision for manipulating specific (e.g., transparent, highly-specular, underwater) objects. Fortunately, bird vision has visual acuity superior to other vertebrates and accurately detects moving underwater prey from an aerial perspective. Specifically, research for correcting the error in depth perception caused by refraction in the presence of air-water interface will be introduced together with its application to robot manipulation.
language of the presentation: English
 
DODI FITRA CHANDRA 1351205: M, 1回目発表
title: Relation Between Centrality of Genes in PPI Network With Their Essentiality
abstract: In Biological system, most of cellular functions are direct result of interaction of among proteins. Therefore Protein-protein Interactions (PPIs) can help us to understand the complexity within an organism. Protein-protein interaction play a key role in the completion of cellular functions and usually correlate to each other in the form of protein-protein network. In this research we will find out the relations between centrality measures and Essentiality of the genes/protein, what type of centrality measures are more appropriate to predict essential genes.
language of the presentation: English
 
葭本 匠 1051127: M, 1回目発表
title:Modeling of fingertip deformation for estimation of fingertip forces
abstract:It is known that haptic information from fingertips is important for human dexterous manipulation. To clarify human ability of manipulation, there are many studies about measuring such kind of haptic information during object manipulation. First, in this presentation, I introduce two related studies about a development of glove type tactile sensor and a fingertip force estimation method using a camera based on fingertip deformation. Next, I explain the proposal estimation method including measuring devices and calibration systems for fingertip lateral deformation.
language of the presentation:Japanese
発表題目:指先力推定に向けた指腹部変形のモデル化
発表概要:人間の器用な物体操作において,捜査中に指先から得られる触覚情報は重要な役割を占めていることが知られている.人間の物体操作スキルを理解するためには,物体操作中にこのような触覚情報を計測する必要があり,様々な研究が行われている.本発表では,関連研究としてグローブ型の接触圧センサの開発およびカメラを用いた指先変形に基づく指先力推定の2つの研究例を紹介し,本研究の提案手法を説明する.また,提案手法で用いる指先変形計測デバイスとキャリブレーションシステムについて紹介する.