权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Security Enhancement and Power Reduction of Networks based on Machine learning Approach with VLSI Technology

基于机器学习方法和 VLSI 技术的网络安全增强和功耗降低

基本信息

批准号：
18500048
负责人：
ABE Koki
金额：
$ 2.38万
依托单位：
The University of Electro-Communications
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2006
资助国家：
日本
起止时间：
2006 至 2007
项目状态：
已结题

项目摘要

I conducted research from various aspects to show that machine learning approaches combined with VLSI technology are effective for security enhancement, performance improvement, and power reduction of networks and their components. Research achievements are summarized as follows:1. Security enhancement of networks utilizing VLSI technologyI proposed an architecture for a fast and accurate signature-based intrusion detection system (IDS) where hardware is used at upstream steps requiring high speed, and software is used at downstream steps requiring high accuracy. I verified the effect by a simulation experiment.2. Power reduction and resource savings of networksI proposed a new method for decreasing the power consumption of wireless networks by predicting the probability of packet collisions. I proposed a congestion control scheme which reduces the congestion of Internet links by machine learning approach to save the network resource and improve the fairness among data flows.3. Improving the performance of processor based on machine learning approachesI improved the performance of CPU as an component of networks using a low-cost and highly accurate branch prediction scheme based on a machine learning approach.4. Enhancing the security of components using VLSI technologyI proposed a countermeasure against differential power analysis attacks to cryptosystems. The countermeasure is suitable for implementation on VLSI. I verified its effectiveness by experiments5. Improvements of security, performance, quality, and power consumption of multimedia network communicationsI proposed an architecture of two dimensional discrete wavelet transform (2D-DWT) to improve the performance and power savings, and utilized the architecture for image data authentication and tamper proofing. I improved the 2D-DWT algorithm to improve the performance and quality of tiling and restoring processes for image data.

我从多个方面进行了研究，证明了机器学习方法与VLSI技术相结合对增强网络及其组件的安全性、提高性能和降低功耗是有效的。1.基于VLSI技术的网络安全增强我提出了一种快速、准确的基于签名的入侵检测系统的体系结构，其中上游步骤使用硬件，下游步骤使用软件，要求高精度。通过仿真实验验证了该方法的有效性。提出了一种通过预测分组冲突概率来降低无线网络功耗的新方法。提出了一种拥塞控制方案，通过机器学习的方法减少了互联网链路的拥塞，节约了网络资源，提高了数据流之间的公平性。基于机器学习方法的处理器性能改进I采用了一种基于机器学习方法的低成本、高精度的分支预测方案，提高了作为网络组件的CPU的性能。利用VLSI技术提高组件的安全性我提出了一种针对密码系统的差分功耗分析攻击的对策。该对策适合在VLSI上实现。我通过实验验证了它的有效性。提高多媒体网络通信的安全性、性能、质量和功耗我提出了一种二维离散小波变换(2D-DWT)结构来提高性能和节能，并将该结构用于图像数据认证和防篡改。我改进了2D-DWT算法，提高了图像数据平铺和恢复过程的性能和质量。