| 近藤 嵩之 | M, 2回目発表 | 情報セキュリティ工学 | 林 優一, 岡田 実, 藤川 和利, 藤本 大介, Kim Youngwoo |
|
title: Enhancing the Quality of Restored Leak Sound Information by Focusing on Multiple Modulation Signals in Reflected Electromagnetic Leakage
abstract: Threats have been reported in which electromagnetic waves are intentionally irradiated onto audio devices, enabling the acquisition of sound information through reflected waves. When observing such leaked information from a distance, it is necessary to suppress the impact of overlapped noise during its propagation process. On the other hand, sound information typically does not repeatedly leak the same information, making it difficult to immediately apply the noise reduction methods based on averaging traditionally used in electromagnetic information security. This study focuses on physical quantities that can capture leaked information at the same time, aiming to enhance signal reception sensitivity, improve the signal-to-noise ratio, and reduce noise's influence on information acquisition. language of the presentation: Japanese | |||
| SHIBLY KABID HASSAN | D, 中間発表 | サイバーレジリエンス構成学 | 門林 雄基, 林 優一, 和田 隆広, 妙中 雄三 |
|
Title: Surviving the Cyber Onslaught: Strategies for Enhancing Connected and Autonomous Vehicle Network Resilience
Abstract: As the complexity and connectivity of Connected and Autonomous Vehicles (CAVs) continue to increase, so does the importance of ensuring the security of their in-vehicle networks. Our research presents a comprehensive approach to safeguarding these networks against various threats, demonstrating the effectiveness of leveraging advanced machine learning techniques. Our proposed defense mechanisms address multiple vulnerabilities, starting with the security risks inherent in deep learning-based driving models. We introduce a robust autoencoder-based system capable of mitigating adversarial attacks, yielding performance that surpasses conventional defense strategies. Simultaneously, we tackle the security weaknesses within the Controller Area Network (CAN) bus, a crucial communication component in modern vehicles. We propose a Federated Learning-based Intrusion Detection System that conducts secure training procedures without requiring data sharing and effectively detects CAN bus attacks. In response to the surge of in-vehicle network traffic due to the proliferation of embedded devices, we propose an intrusion detection system for Automotive Ethernet in In-Vehicle Networks (IVN). This system employs semi-supervised learning, enhancing the algorithm's ability to discern harmful activities in the network traffic, and thus improving the security of the IVN. In conclusion, our research proposes a multi-faceted, machine learning-based approach to secure in-vehicle networks in Connected and Autonomous Vehicles. The security measures we present are not only effective but also adaptable to the evolving landscape of vehicle technologies, promising to enhance the safety and reliability of future transportation systems. Language of Presentation: English | |||
| TONGIAM PACHARAWAN | 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, Md.Delwar HOSSAIN |
|
Title: Exploring the Potential of Deep Learning for Malware Detection in Android Mobile Banking: A Case Study of Thailand's Financial Sector
Abstract: One of Thailand's most severe cyber threats occurred when attackers launched a series of cyberattacks on the financial sectors. The users experienced blackmail, phishing, and personal data exposure to the public, stealing money and even direct financial theft. The patterns always changed and varied based on the financial sectors’ structures and the nature of the clients. This research is inspired by the growing cyber threats faced by Thailand's financial sector, particularly in mobile banking, a concern that has come to prominence due to a series of high-profile cyberattacks. Such attacks have caused significant losses, both financial and reputation, to users and financial institutions. The study's objectives are to investigate the effectiveness of Deep Learning algorithms in detecting malware in the Android mobile banking system, and to compare the performance and accuracy of these algorithms. Language of the presentation: English | |||
| ISLAM MD SIHABUL | M, 1回目発表 | ディペンダブルシステム学 | 井上 美智子, 中島 康彦, 江口 僚太 |
|
Title: Enhancing the Reliability of Memristor Crossbar-Based Neuromorphic Computing System
Abstract: A Neuromorphic Computing System (NCS) may be enabled to acquire knowledge and carry out a task independently through interaction with its surrounding environment. A combination of such chips with CMOS-based processors may be able to address a number of issues that are being dealt with by artificial intelligence (AI) systems. While CMOS-based architectures have been developed to enhance the computational efficiency of AI applications, core operations such as matrix multiplication and convolution still heavily depend on CMOS-based multiply-accumulate units. The von Neumann bottleneck poses inherent limitations to these units, hindering overall performance. Luckily, among several emerging memory devices, memristor devices have the inherent capability to directly execute vector-matrix multiplication by leveraging Ohm's law and Kirchhoff's law. This can be achieved by utilizing an array of memristors in a crossbar architecture. However, the presence of stuck-at faults (SAFs) in memristor devices considerably undermines the computational accuracy of NCS. In our research, we aim to propose a reliable design for memristor crossbar based NCS under the consideration of utilizing the dropout technique to enhance reliability. language of the presentation: English | |||