| 砂田 翼 | M, 2回目発表 | ソフトウェア工学 | 松本 健一, | 飯田 元, | 石尾 隆, | 畑 秀明, | Kula Raula Gaikovina |
|
title: Towards Cheating Detection in Programming Tests
abstract: In company recruitment activities, an increasing number of companies use programming tests to evaluate the technical capabilities of candidates. However, cheating is performed in the test, and there is a problem that the candidate's technical ability cannot be evaluated correctly. The purpose of this study is to detect cheating behavior in coding tests. we propose a method for detecting the similarity of source code by using the source code obtained in the actual company coding test and detecting cheat from the similar source code. language of the presentation: Japanese | |||||||
| 森田 大夢 | M, 2回目発表 | ソフトウェア工学 | 松本 健一, | 飯田 元, | 石尾 隆, | Kula Raula Gaikovina | |
|
title: Analyzing the Difficulty of Compiler Errors for Novice Programmers
abstract : Identifying and correcting programming errors is one of the challenging task facing the novice programmer. Educators can make the quality of programming education more effective by understanding how difficult individual error types are for students. In this study, we examined what type of errors do students find difficult. Therefore, we collected error messages generated through the NAIST programming exercise class. Moreover, we examined the probability that each error will occur in the first and second half of the class and types of errors that take a long time to resolve. As a result, I detected some difficult to solve errors. language of the presentation:Japanese | |||||||
| 深澤 佑樹 | M, 2回目発表 | ソフトウェア設計学 | 飯田 元, | 松本 健一, | 市川 昊平, | 髙橋 慧智 | |
|
title: Method name recommendation using seq2seq model based on method call relationship
abstract: Understanding the source code is important in software development. Proper method names are one of the key factors in understanding the source code. However, developers often name inappropriate methods. Along with this, various methods have been proposed to recommend method names. However, the problem is that the accuracy of method name recommendation is low. In this study, we propose a method name recommendation method using the seq2seq model based on the method call relationship, with the aim of recommending method names with high accuracy. In this presentation, the proposed method will be explained, and the evaluation experiment will be the future work. language of the presentation: Japanese | |||||||
| 千田 将也 | M, 2回目発表 | ソフトウェア設計学 | 飯田 元☆, | 松本 健一, | 片平 真史（客員教授）, | 石濱 直樹（客員准教授）, | 高井 利憲（客員准教授） |
|
title: Towards improving transferability of adversarial attacks against regression models
abstract: Recently, machine learning has been attracting attention as a core technology for autonomous vehicles. For example, autonomous steering control using a machine learning model with camera images is widely studied. On the other hand, vulnerabilities in machine learning to adversarial attacks have been reported. Since "connected cars", that communicate with the outside world, are expected to become more common, security for machine learning is becoming an issue. Among the studies on adversarial attacks against machine learning, a method of generating adversarial example based on Generative Adversarial Network (GAN) has been proposed. The property of an adversarial attack that an adversarial example generated under the assumption of an architecture of prediction models is also effective for another prediction model with a different architecture is called transferability. Here, it is reported that for regression models there is little transferability whereas for classification models transferability is sufficiently high. In this study, we propose a method to generate adversarial example for multiple architecture models using GAN in regression cases with the aim of improving the transferability. As a case study, we evaluate the effectiveness of the proposed method against predicting the steering angle of an autonomous vehicle using camera images. In this presentation, we will explain the proposed method, and the evaluation through the case study is left for the future work. language of the presentation: Japanese 発表題目: 回帰モデルに対する敵対的攻撃の転用可能性の向上手法 発表概要: 近年，自動運転車のコア技術として機械学習が注目されている．例えば，カメラ画像を用いた機械学習モデルによる自律的ステアリング制御の研究開発が行われている．一方で，機械学習では敵対的攻撃に対する脆弱性が報告されており，さらに，自動車は今後外部との通信が行われるコネクテッドカーが一般的になることが予想され，セキュリティ上の脅威が課題となっている．機械学習に対する敵対的攻撃方法の研究として，敵対的生成ネットワーク（GAN: Generative Adversarial Network）を応用して敵対的サンプルを生成する手法が提案されている．ここで，あるアーキテクチャの予測モデルを仮定して生成された敵対的サンプルが，異なるアーキテクチャの予測モデルに対しても有効である敵対的攻撃の性質を転用可能性という．先行研究において，回帰モデルでは転用可能性が低いことが報告されている一方で，分類モデルでは転用可能性が十分に高いことが示されている．本研究では，回帰モデルにおける敵対的攻撃の転用可能性を向上させることを目的に，GANを用いた複数の回帰モデルに対する敵対的サンプルを生成する手法を提案する．ケーススタディとして，カメラ画像を用いた自動運転車のステアリング角度予測に対し，提案手法の有効性を確認する．今回の発表では提案手法の説明を行い，ケーススタディによる評価に関しては今後の作業とする． | |||||||
| 大須賀 彩希 | D, 中間発表 | 情報セキュリティ工学 | 林 優一, | 岡田 実, | 中島 康彦, | 藤本 大介, | Ingrid Verbauwhede (KU LEUVEN) |
|
title: Elucidation of the security degradation mechanism of the true random number generator by physical attack
abstract: True random number generators (TRNGs) based on ring oscillators (ROs) are employed in many devices because they can be constructed by digital circuits. If an RO-based TRNG is attacked and its security is compromised, many systems are affected. In this study, I aim to elucidate the mechanism that causes randomness degradation of TRNG using RO due to intentional electromagnetic interference and electromagnetic information leakage, and to develop countermeasures. Specifically, I will evaluate the source of the leakage that causes electromagnetic information leakage, and create a simulation model to elucidate the mechanism of randomness degradation of TRNG due to disturbance waves. Using a simulation model, I will evaluate the impact of TRNG on attacks on both immunity and emissions, and investigate countermeasure technologies. language of the presentation: Japanese | |||||||
| 鍛治 秀伍 | D, 中間発表 | 情報セキュリティ工学 | 林 優一, | 岡田 実, | 井上 美智子, | 藤本 大介, | Youngwoo Kim |
|
title: Development of Trojan-free Hardware Design Technique for the Threat of Electromagnetic Information Leakage
abstract: The threats of information leakage through unintentional electromagnetic (EM) emissions from electronic devices have been reported. This threat was targeted on devices with a high intensity of EM emission, while devices with a low intensity of EM emission were out of the target of the threat. On the other hand, there are reports of Hardware Trojan (HT) that can be implemented on PCBs of devices or lines among devices; one of which is an attack increasing the intensity of EM emission containing internal signals using HT and the continuous wave. Since the HT used in this attack is a simple structure, there is a possibility that it is implemented in actual devices. This study shows the possibility of EM information leakage induced by HT and continuous wave and clarifies the leakage mechanism by estimating the module structure that induces the leakage. Then, a Trojan-free hardware design technique will be proposed. language of the presentation: Japanese | |||||||
| 和田 慎平 | D, 中間発表 | 情報セキュリティ工学 | 林 優一, | 岡田 実, | 安本 慶一, | 藤本 大介, | Youngwoo Kim |
|
title: Design Methodology for Cryptographic Devices Resistant to Electromagnetic Information Leakage
abstract: In recent years, electromagnetic analysis (EMA) attacks by analyzing the electromagnetic (EM) radiation at a distance from the module have been reported and devices can be the vulnerable to the EMA attacks even if the countermeasure such as a tamper-resistant design is applied to the module. In this research, the objective is to establish a design methodology for cryptographic devices resistant to EM information leakage. This research consists of (1) a proposal of an efficient EMA based on the side-channel measurement method focusing on the design pattern of the printed circuit board (PCB), (2) investigation of the mechanism of the EM radiation containing the secret key information, (3) development of a simulation method to evaluate EM information leakage at the design stage of cryptographic devices and (4) discussion countermeasures against EM information leakage. language of the presentation: Japanese | |||||||