| 宇野 真純 | 1551018: M, 2回目発表 | 情報基盤システム学 | 藤川 和利,安本 慶一,松本 健一,新井 イスマイル,猪俣 敦夫 |
|
title:A RAT detection method by using packet emtropy on early stage
abstract: Confidential information leak accidents by targetted attacks has becomes a serious social issue. In targetted attacks, Remote Access Trojan/tool (RAT) is heavily used. To minimize damage by the attack, it is important to detect the RAT activity on intrusion stage. The detection the RAT communication is getting more and more difficult to detect with technology advances. Previous studies can detect limits protocols on specific conditions but they cannot cover advanced RAT. In this study, we aim making a general method detecting on earlystage of RAT communication and purpose a method to use packet entropy of the communication. language of the presentation: japanese | |||
| 上村 康輔 | 1551035: M, 2回目発表 | ソフトウェア工学 | 松本 健一,藤川 和利,門田 暁人,畑 秀明 |
|
title: Software Reuse Detection Using Binary Program Split Compression
abstract: Open source software (OSS) is utilizable and obtainable anyone. However, the source code theft has happened to exploit it. If the developer finds the suspicious commercial software, they can't see the source codes because only available in binary. Therefore, the proof that software is stolen is difficult. Since there're some plagiarism detection method for binary program, they still have problems like requiring complex task and/or hard to detect small plagiarism. In this study, we propose a novel method which can solve these problems. Our method can easily detect even small plagiarism by splitting binary program into small pieces, compressing them and comparing the size of compressed pieces. | |||
| 弥谷 圭朗 | 1551111: M, 2回目発表 | ソフトウェア工学 | 松本 健一,藤川 和利,楫 勇一 |
|
title: Improvement of Winternitz one time signature and its security proof
abstract: One-Time Signature (OTS) is a light-weighted and quantum immune alternative of digital signatures. OTS is available only once, and typically constructed by hash functions only. Winternitz OTS is a widely recognized OTS, but it suffers for long signature size. An improvement of Winternitz OTS is investigated in this study, by em- ploying a novel constant-sum fingerprint function which helps reducing the size of the signature and the complexity for signature verification, and further by using a random nonce that contributes to reduce the complexities for key generation and signature computation. language of the presentation : Japanese 発表題目：改良型 Winternitz one time 署名の提案と安全性評価 発表概要：One Time 署名 (OTS) とは,ハッシュ関数のみを用いて実現される使い捨て型の電子署名技術てあり,比 較的計算量か小さく,量子計算機の出現によっても危殆化しない等の特長を持つ.代表的な OTS として Winternitz 署名か知られているか,十分な強度を確保したまま署名サイスを小さくすることか難しいという問題かある.そこて 本研究ては,定数和指紋関数と呼はれる特殊な fingerprint 関数を導入して署名サイスおよひ署名検証の計算量を削減 し,さらに,署名生成に乱数要素を利用することて,鍵生成およひ署名作成の計算量も削減する手法を提案する | |||
| HOLLAND MATTHEW JAMES | 1561025: D, 中間発表 | 数理情報学 | 池田 和司,笠原 正治,佐々木 博昭 |
|
Title: Stable learning via robust objectives Abstract: The theoretical foundations of machine learning are centred around questions of the optimality and fundamental limitations of statistical procedures. Formal insights form a basis for the principled development of reliable algorithms with well-understood performance guarantees. In this presentation, we look at the task of predicting real values given multi-dimensional patterns, namely the classical regression problem, and highlight performance issues underlying well-known classes of algorithms. In this work, we put forward a general method of algorithm design, which deviates from the traditional risk-centric paradigm in order to build computable objective functions which provide stabler targets. Novel formal analysis of statistical properties, computational routines, and controlled numerical simulations form the bulk of our results thus far. In today's talk we shall give a broad overview of our progress thus far and a short-term look ahead. 発表概要： 機械学習と呼ばれる分野の理論的な基盤は、統計的推定方法の最適性やパフォーマンスの限界をめぐる疑問に深く根付いている。理論上の明瞭で的確な知見を踏まえてこそ、信頼性の高い学習アルゴリズムを体系的に構築する枠組みが見えてくる。本日の発表では、多次元パターンから実数値を取る応答を予測するという回帰問題に主眼を置く。まずは著名な従来の推定方法の弱点を指摘し、本研究の動機づけを固める。本研究では、伝統的なリスク最小化の枠組みから多少それて、安定性の優れた推定量が設計しやすいように、推定問題の「ターゲット」たるものを自ら設計することから出発する。この議論を経て新しいアルゴリズムの導出と、その統計的性質の解析、実用性の高い計算法の提案や数値実験による性能評価といった成果を報告し、今後に向けての方針を述べる。 Language of the presentation: Japanese | |||
| 藤井 博晃 | 1551089: M, 1回目発表 | 数理情報学 | 池田 和司,笠原 正治,久保 孝富 |
| 土田 将司 | 1551063: M, 2回目発表 | ディペンダブルシステム学 | 井上 美智子,笠原 正治,大下 福仁 |
|
title: Gathering of mobile agents which has Byzantine failue tolerance with authenticated whiteboards
abstract:We propose an algorithm for the gathering problem of mobile agents in Byzantine environments. The proposed algorithm can make all correct agents to meet at a single node in \(O(fm)\) time (\(m\) is the number of edges) under the assumption that each agent has unique ID and behaves synchronously, each node is equipped with an authenticated whiteboard, and at most \(f\) Byzantine agents exist. Since the existing algorithm achieves gathering without a whiteboard in \(\tilde O(n^9\lambda)\) time, where \(n\) is the number of nodes and \(\lambda\) is the length of the longest ID, our algorithm shows a whiteboard can significantly reduce the time for the gathering problem in Byzantine environments. language of the presentation:Japanese | |||
| 小松 巡 | 1451201: M, 1回目発表 | ディペンダブルシステム学 | 井上 美智子 |
|
title:(Paper introduction)A new self-stabilizing maximal matching algorithm
abstract:A matching in an undirected graph is a subset of edges in which no pair of edges are adjacent. A matching M is maximal if no proper superset of M is also a matching. This paper presents a self-stabilizing algorithm for this problem that has the same time complexity as the previous best algorithms for the sequential adversarial, the distributed fair, and the synchronous daemon. language of the presentation:Japanese | |||