コロキアムB発表

日時: 07月18日 (Fri) 2限目(11:00-12:30)


会場: L1

司会: 坂井 優介
名越 遼 M, 2回目発表 情報セキュリティ工学 林 優一, 岡田 実, 向川 康博, 藤本 大介, 鍛治 秀伍
title: A Study on Inducing Electromagnetic Information Leakage by Irradiating Multiple Frequencies
abstract: Unintentional electromagnetic (EM) emissions are generated by fluctuations of electrical signals inside electronic devices. Security threats have been reported that internal device information is leaked by capturing and analyzing these emissions. On the other hand, devices with weak EM emissions have been considered outside the scope of such threats, as it is difficult to capture EM emissions. Therefore, a method has been proposed that irradiates EM waves from outside the device to enhance emission strength. However, this method has a problem in obtaining information from EM waves due to self-interference caused by the superimposition of irradiation EM waves on the radiated EM waves, as both have the same frequency. This study proposes a method that causes EM information leakage with a different frequency from the irradiated EM waves, focusing on frequency conversion that occurs when multiple frequencies propagate to a nonlinear circuit element implemented inside the device. This method mitigates the effect of self-interference and enhances the obtainability of information from EM waves.
language of the presentation: Japanese

 
松本 匠平 M, 2回目発表 情報セキュリティ工学 林 優一, 岡田 実, 向川 康博, 藤本 大介
title: A Study on Electromagnetic Information Leakage from Display Devices Focusing on Character Reconstruction through Pixel-level Analysis
abstract: Display devices unintentionally emit electromagnetic (EM) radiation during operation, allowing attackers to eavesdrop on screen contents. In particular, High-Definition Multimedia Interface (HDMI) cables emit strong EM radiation, leading to information leakage over a distance from target devices. On HDMI cables, however, multiple signal patterns can correspond to the same color. As a result, conventional reconstruction methods often reproduce a single color as multiple colors, blurring the boundary between foreground and background. This issue has been considered a major issue in reconstructing small characters, such as sensitive information. This study proposes a character reconstruction method based on binary pixel classification, focusing on the common characteristic of screens that consist only of two colors, foreground and background. The proposed method indicates that characters can be reconstructed using EM radiation leaked from target devices.
language of the presentation: Japanese
 
金谷 祐輝 M, 2回目発表 情報セキュリティ工学 林 優一, 中島 康彦, 岡田 実, 藤本 大介
title: A Study on Fault Sensitivity Estimation Using Side-channel Information
abstract: Fault sensitivity analysis, one of the fault injection attacks, recovers the secret key by focusing on the input plaintext dependence of the clock interval at which a fault begins to occur (fault sensitivity) when the clock interval supplied to the cryptographic module is gradually decreased. In contrast, a countermeasure method has been proposed to hide the plaintext dependence of the fault sensitivity by stopping the output of ciphertext before the clock interval reaches the fault sensitivity. However, since the countermeasure works just before the ciphertext is output, the encryption process is still being executed, and the side-channel information generated during the process may be used to estimate the fault sensitivity. Although previous research has shown that a weak positive correlation can be obtained between side-channel information and fault sensitivity. By contrast, the secret key has not been recovered due to estimation errors caused by the plaintext dependence of side-channel variation during faults and measurement noise. This study proposes a quantitative evaluation method for the impact of estimation error on secret key recovery performance. This method provides an upper bound on the acceptable estimation error when recovering secret keys.
language of the presentation: Japanese
 
小泉 理久 M, 2回目発表 情報セキュリティ工学 林 優一, 中島 康彦, 岡田 実, 藤本 大介
title: A Study on Energy-efficient Hardware Implementation of Number Theoretic Transform for CRYSTALS-Kyber
abstract: The advent of large-scale quantum computers poses a serious threat to classical public-key cryptosystems such as RSA and elliptic curve cryptography. In response, international efforts are underway to standardize post-quantum cryptography, and the National Institute of Standards and Technology (NIST) has selected CRYSTALS-Kyber as a key encapsulation mechanism. In CRYSTALS-Kyber, modular multiplication in the number theoretic transform is the main computational bottleneck. Particularly, improving energy efficiency is an important issue for the widespread adoption of systems that require low power consumption. Most existing FPGA implementations focus on minimizing circuit area and use multipliers based on DSP blocks. In contrast, future ASIC implementations will support pipelining and flexible resource allocation, requiring different design strategies. This study presents a DSP-independent modular multiplier architecture, implemented and evaluated on an FPGA, with a focus on optimizing energy efficiency. As a result, the proposed design achieved a throughput improvement over conventional multipliers and reduced the area-delay product by 19% on Artix-7 and 23% on Kintex-7.
language of the presentation: Japanese
 

日時: 07月18日 (Fri) 2限目(11:00-12:30)


会場: L2

司会: Christopher Wiraatmaja
大西 一誉 D, 中間発表 ソーシャル・コンピューティング(ロボット対話知能) 荒牧 英治☆, 吉野 幸一郎, 河野 誠也
title: Modeling Turn-Taking Speed and Speaker Characteristics
abstract: Modeling turn-taking speed while considering speaker characteristics and the relationships between speakers is essential for realizing dialogue systems capable of natural interactions. In this study, we focused on dialogue participants' roles, relationships, and personality, analyzing and modeling turn-taking speeds observed in real conversations. The analysis confirmed that the expression of these attributes—role, relationship, and personality—is closely associated with turn-taking speed. Based on these findings, we constructed a model that predicts the distribution of turn-taking speeds according to each attribute using a gamma distribution. Evaluation results demonstrated that appropriate parameter fitting to the three-parameter gamma distribution enables effective modeling of turn-taking speeds based on participants' roles, relationships, and characteristics.
language of the presentation: Japanese
発表題目: ターンテイキング速度と話者特性のモデリング
発表概要: 話者特性および話者間の関係性を考慮したターンテイキング速度のモデリングは、より自然な対話を実現する対話システムの構築に不可欠である。本研究では、対話参加者の役割、関係性、そしてパーソナリティに着目し、実際の会話において観測されたターンテイキング速度を分析・モデル化を行なった。分析の結果、これらの属性(役割、関係性、パーソナリティ)の表出がターンテイキング速度と密接に関連していることが確認された。そこで得られた知見に基づき、各属性ごとのターンテイキング速度の分布をガンマ分布を用いて予測するモデルを構築した。評価実験の結果、ガンマ分布への適切なパラメータフィッティングにより、参加者の役割・関係性・特性に応じたターンテイキング速度を効果的にモデリングできることが示された。
 

日時: 07月18日 (Fri) 2限目(11:00-12:30)


会場: L3

司会: Fan Youmei
古谷 大介 M, 2回目発表 光メディアインタフェース 向川 康博, 和田 隆広, 舩冨 卓哉, 藤村 友貴, 北野 和哉
title: Restoration of lost parts of ancient tomb murals and estimation of pigment thickness using spectral images
abstract: Murals of ancient tombs may be lost due to deterioration over time, and it is historically valuable to know the state of the murals when they were painted. In this study, we attempted two methods to restore the lost parts of the murals and estimate the thickness of the pigments using spectral images. First, we performed supervised learning using the parts that can be distinguished by the human eye as training data, and were able to confirm the presence or absence of pigments in the lost parts. Second, we proposed a method to estimate the thickness of the pigments without depending on how the training data was given. Existing methods using physics-based autoencoders have problems such as the requirement that the underlying rocks be in an ideal state where they are uniform, and the inability to interpret the latent space with a simple autoencoder. Therefore, we aim to estimate the thickness of the pigments without a teacher by using a network that combines an autoencoder and a physical formula.
language of the presentation: Japanese
発表題目: 分光画像を用いた古墳壁画の消失部分の復元と顔料の厚み推定
発表概要:古墳壁画は経年劣化による喪失の可能性があり、描かれた当時の状態を知ることは歴史的価値がある。本研究では、分光画像を用いて壁画の消失部分の復元と顔料の厚みを推定するために、2種類の方法を試みた。1つ目として、人間の目で判別できる部分を教師データとして教師あり学習を行った結果、消失部分の顔料の有無を確認することができた。2つ目に、識別結果が教師データの与え方に依存せずに顔料の厚みを推定する方法を提案した。既存の物理ベースオートエンコーダを使った手法では下地の岩石が均一であるという理想的な状態でなければならないことや、単純なオートエンコーダでは潜在空間を解釈することができないといった問題があった。そこで、オートエンコーダと物理的な数式を組み合わせたネットワークを用い、教師なしで顔料の厚みを推定することを目指す。
 
三好 湧晴 M, 2回目発表 ヒューマンロボティクス 和田 隆広, 松原 崇充, 劉 海龍, 織田 泰彰, 本司 澄空
title:Sequential estimation of states and parameters of a tendon-driven robot finger with soft joints
abstract:Soft robots made from polymer materials can undergo significant deformation, and soft grippers that utilize this feature can grasp objects stably in accordance with their shapes. On the other hand, there is variability in movement due to the nonlinearity of the materials that create softness and external factors such as the surrounding environment. To mathematically capture such variability, models incorporating random variables under the assumption that parameters within the model change with each trial have been proposed. However, these parameters are also believed to change during motion. In this presentation, we propose a method for stable parameter estimation by leveraging insights from prior research and simulations.
language of the presentation:Japanese
発表題目:柔軟関節を有する腱駆動型ロボットフィンガの状態およびパラメータ逐次推定
発表概要:高分子材料などで作成されるソフトロボットは大きな変形が可能であり, この特徴を活かしたソフトグリッパは物体の形状に沿った安定した把持を行うことができる. 一方で柔らかさを生み出す材料の非線形性や周辺環境などの外的要因により運動のばらつきが存在する. こうしたばらつきを数学的に捉えるため,モデル内のパラメータが試行ごとに変化するという仮定のもとで確率変数を導入したモデルが提案されているが, これらパラメータは運動中にも変化していると考えらえる.そこで本発表では,先行研究の知見やシミュレーションを活用し, 安定したパラメータ推定が可能な手法を提案する.
 
森本 奏多 M, 2回目発表 ヒューマンロボティクス 和田 隆広, 松原 崇充, 劉 海龍, 織田 泰彰, 本司 澄空
title: Path Generation and Tracking Control for Teleoperated Robots Using Wave Variable
abstract: In Teleoperation of mobile robots, considering communication delay is critically important for ensuring operational stability. As a solution to this issue, a method based on passivity has been proposed in prior research. In this method, the operator's steering input and the robot's angular velocity output are converted into wave variables, which are then transmitted and received to maintain passivity over a delayed communication channel. However, when this method was applied to an actual differential-drive mobile robot, experimental results revealed instability in the robot’s velocity output. To address this issue, this research proposes an approach in which the operator does not directly provide control inputs to the robot, but instead inputs a desired trajectory. This trajectory is then transformed into wave variables and transmitted, allowing the robot to autonomously follow the received path. Using this system, we conduct Teleoperation experiments and evaluate its effectiveness.
language of the presentation: Japanese
発表題目: 遠隔操作ロボットのための波動変数変換を用いた経路生成と追従制御
発表概要: 車輛の遠隔操作において通信遅延を考慮することは,操作の安定性を確保するうえで極めて重要である.この課題に対する先行研究として,受動性に基づく遠隔操作手法が提案されている.この手法では,操作者の操舵入力および車両の角速度出力を波動変数に変換し,これを介して送受信をおこなうことで,遅延を有する通信路における受動性の維持を可能にしている.しかしながら,この手法を実際の差動二輪ロボットに適用した場合、ロボットの速度出力に不安定性が生じることが実験的に確認された.そこで本研究では,この問題に対処するために,操作者はロボットの制御入力を直接与えるのではなく,走行経路を入力し,それを波動変数に変換して伝送することで,車輛が受信した経路に従って自律的に追従する手法を提案する.本システムを用いて実際に遠隔操作を行い,その有効性について検証を行う.