| 上谷 仁亮 | M, 1回目発表 | コンピューティング・アーキテクチャ | 中島 康彦, 林 優一, 張 任遠, KAN Yirong, PHAM HOAI LUAN |
|
title: Accelerated 3D real-time ionospheric tomography
abstract: In the region known as the ionosphere, situated above Earth, occurrences (disturbances) transpire wherein the electron density undergoes rapid changes within a brief temporal span. These ionospheric disturbances pose challenges as they have the potential to disrupt critical infrastructure, including satellite communications. Numerous unresolved facets of these disturbance phenomena persist, necessitating continual observation. Consequently, this study centers on ionospheric tomography observation—an extension of GNSS-TEC observation (TEC: Total Electron Content), a technique utilizing radio waves from positioning satellites like GPS. However, tomographic observations demand substantial computational resources, impeding the attainment of desired real-time performance. To address this, our study seeks to enhance the speed of tomography observation via two approaches: (1) hardware enhancements involving technologies such as GPGPUs and computer architectures like the Coarse-Grained Reconfigurable Architecture (CGRA), as proposed by our laboratory, and (2) algorithmic refinements implemented through software. language of the presentation: Japanese | |||
| NDALAMA FESTUS EDWARD | M, 1回目発表 | サイバーレジリエンス構成学 | 門林 雄基, 林 優一, 妙中 雄三, HOSSAIN, Md Delwar |
|
Title: Behavioral Biometrics for Continuous Authentication on Internet Banking for Smartphone Users on a Web Browser: A Deep Learning Perspective Abstract: With the increasing reliance on Internet banking services, we evaluate the feasibility and effectiveness of incorporating Behavioral Biometrics for continuous authentication in the context of Internet banking accessed through web browsers on smartphones. Utilizing a deep learning approach, the research aims to analyze user behavior patterns, including touch gestures and typing dynamics, with a focus on smartphone-specific interactions. The study, in the planning stages, envisions establishing a reliable and non-intrusive authentication mechanism. Emphasizing adaptability to evolving user interactions, this research anticipates providing valuable insights for financial institutions seeking advanced, user-friendly security measures. Rigorous evaluation against a spectrum of attack scenarios, including impersonation attacks, replay attacks, data breaches, phishing attempts, and machine learning adversary attacks, will set our research apart and contribute to preventing unauthorized access. The research findings are expected to enhance the user experience, presenting a dynamic and effective security solution for Internet banking transactions. Language of presentation: ENGLISH | |||
| 米倉 未樹 | M, 1回目発表 | ソフトウェア設計学 | 飯田 元, 松本 健一, 市川 昊平, 平尾 俊貴, 柏 祐太郎 |
|
title: Toward Proposing a Context-Aware Self-Admitted Technical Debt Detection Tool
abstract: Self-Admitted Technical Debt (SATD) refers to defects and issues that exist in the code that need to be resolved, of which the developer is aware of the issues and has embedded them in the code. For example, a developer can write SATD comments in the code to let the team know that the current implementation is not optimal and needs future maintenance. In recent years, various SATD detection methods have been proposed to facilitate analysis of SATD. However, there have been scattered cases where comments describing the processing of source code are mis-detected as SATD. In this study, we attempt to improve the accuracy of SATD detection by considering the context in the source code to prevent false positives. Specifically, we use CodeBERT to learn comments and the source code immediately below as input. language of the presentation: Japanese 発表題目: コンテキストを考慮したSelf-Admitted Technical Debt検出ツールの提案にむけて 発表概要: Self-Admitted Technical Debt（SATD）とは、コード中に存在する不具合や解消すべき課題のことであり、その中でも開発者が課題を認識した上でコードに埋め込んだものを指す。例えば、開発者はSATDコメントをコード内に記述することで、現在の実装が最適でなく、将来の保守が必要であることをチームに周知することができる。 近年ではSATDを容易に分析するために，様々なSATD検出手法が提案されている．しかし，ソースコードの処理を説明するコメントがSATDとして誤検出される例が散見されている．本研究ではソースコード内のコンテキストを考慮することで誤検出を防止し、SATD検出精度の向上を試みる．具体的には，CodeBERTを用いてコメントと直下のソースコードを入力として学習する． | |||
| 森川 靖仁 | M, 1回目発表 | ソフトウェア設計学 | 飯田 元, 松本 健一, 市川 昊平, 平尾 俊貴, 柏 祐太郎 |
|
title: A token-level inline comment recommendation in modern code review.
abstract: Code review is an important process for ensuring software quality. However, it requires a significant amount of time, with over 40% of reviews reported to take more than a day before receiving initial feedback. In recent years, many methods have been proposed to recommend lines that should be pointed out during the review to support the review. However, because these are line-level recommendations, the granularity is coarse, and there is a problem that the places developers should point out are ambiguous. Therefore, in this study, we propose a method to recommend tokens that should be pointed out using LSTM and Attention. language of the presentation: Jananese 発表題目: モダンコードレビューにおけるトークンレベルのインラインコメント箇所推薦手法の提案 発表概要: コードレビューはソフトウェア品質を確保するため重要なプロセスである．コードレビューは多大な時間を必要とするため，40%以上のレビューが最初のフィードバックを受けるまでに1日以上を費やしていると報告されている．近年ではレビューを支援する目的として，レビュー中に指摘すべき行を推薦する手法が数多く提案されている．しかしながら，行レベルの推薦であるため粒度が荒く，開発者が指摘すべき箇所が曖眛である問題が存在する．そこで本研究ではLSTMとAttentionを用いて指摘されるべきトークンを推薦する手法を提案する. | |||
| 宇恵 勝紀 | D, 中間発表 | 数理情報学 | 池田 和司, 作村 諭一（BS）, 川鍋 一晃（客員教授）, 田中 沙織, 久保 孝富 |
|
title: *** Investigation of a biophysical mathematical model to reproduce the MEG signal ***
abstract: *** There are many mathematical models for resting-state brain activity, and different researchers deal with different models. And there is little debate about which biophysical mathematical model is better for directly understanding the mechanisms that generate brain activity. This may prevent better understanding of brain dynamics. In this presentation, I will discuss the reproducibility of signals of brain activity by the Wilson-Cowan model as a first step for considering an appropriate model. Unlike in the previous doctoral program, I will target the reproduction of MEG signals that indicate early brain activity. Finally, I will show the research that I plan to conduct in the future. *** language of the presentation: *** Japanese *** 発表題目: *** MEG信号を再現する生物物理的な数理モデルの検討 *** 発表概要: *** 安静時脳活動を表す数理モデルは多数存在し、研究者ごとに扱うモデルが異なる。そして、脳活動の発生メカニズムの理解に直接的に役立つ生物物理的な数理モデルについては、どのモデルが良いかについての議論はされていない。このため、脳ダイナミクスへの理解が進まない恐れがある。今回の発表では、適切なモデルの検討の前段階として、Wilson-Cowanモデルによる脳活動の信号の再現性について検討する。博士前期課程とは異なり、早い脳の活動を示すMEG信号を再現目標とする。最後に、今後行う予定の研究内容を示す。 *** | |||