基于轻量级神经网络的白盒攻防研究

针对轻量级神经网络在移动端的安全性缺乏问题,提出模型安全的反防御和二次防御方法。通过在MobileNet v2模型上改造CW算法优化器,采用旋转、添加高斯噪声以及双边滤波器平滑的传统数据增强进行防御,进而体现出反防御效果。在MobileNet v2、ShuffleNet v2和MnasNet模型上,通过改变旋转缩放因子、采用高斯-X噪声混叠模式以及调整双边滤波像素中心点直径对模型进行二次防御。实验结果表明,ASGD优化器对CW算法的反防御能力有很好的提升,同时改进的数据增强方法可以降低对抗样本的鲁棒性,提升模型二次防御能力。     

Differential Fault Attack on GIFT

GIFT,a lightweight block cipher proposed at CHES2017,has been widely cryptanalyzed this years.This paper studies the differential diffusion characteristics of round function of GIFT at first,and proposes a random nibble-based differential fault attack.The key recovery scheme is developed on the statistical properties we found for the differential distribution table of the S-box.A lot of experiments had been done and experimental results show that one round key can be retrieved with an average of 20.24 and 44.96 fault injections for GIFT-64 and GIFT-128 respectively.Further analysis shows that a certain number of fault injections recover most key bits.So we demonstrate an improved fault attack combined with the method of exhaustive search,which shows that the master key can be recovered by performing 216 and 217 computations and injecting 31 and 32 faults on an average for GIFT-64 and GIFT-128 respectively.     

Deriving tabular event-based specifications from goal-oriented requirements models

Goal-oriented methods are increasingly popular for elaborating software requirements. They offer systematic support for incrementally building intentional, structural and operational models of the software and its environment. They also provide various techniques for early analysis, notably, to manage conflicting goals or to anticipate abnormal environment behaviours that prevent goals from being achieved. On the other hand, tabular event-based methods are well-established for specifying operational requirements for control software. They provide sophisticated techniques and tools for late analysis of software behaviour models through simulation, model checking or table exhaustiveness checks. The paper proposes to take the best out of these two worlds to engineer requirements for control software. It presents a technique for deriving event-based specifications, written in the SCR tabular language, from operational specifications built according to the KAOS goal-oriented method. The technique consists of a series of transformation steps each of which resolves semantic, structural or syntactic differences between the KAOS source language and the SCR target language. Some of these steps need human intervention and illustrate the kind of semantic subtleties that need to be taken into account when integrating multiple formalisms. As a result of our technique SCR specifiers may use upstream goal-based processes à la KAOS for the incremental elaboration, early analysis, organization and documentation of their tables, while KAOS modelers may use downstream tables à la SCR for later analysis of the behaviour models derived from goal specifications.     

基于神经网络结构搜索的轻量化网络构建

轻量化网络可解决深度神经网络参数较多、计算量较高、难以部署在计算能力有限的边缘设备上等问题.针对轻量化网络中常用的分组卷积的分组结构问题,文中提出基于神经网络结构搜索的轻量化网络.将不同分组的卷积单元作为搜索空间,使用神经网络结构搜索,得到网络的分组结构和整体架构.同时为了兼顾准确率与计算量,提出循环退火搜索策略,用于解决神经网络结构搜索的多目标优化问题.在数据集上的实验表明,文中网络识别准确率较高,时间复杂度和空间复杂度较低.     

小口径离轴抛物面反射镜挠性支撑设计分析

关于反射镜支撑稳定性优化设计问题,离轴抛物面反射镜作为空间光学遥感器的重要光学元件,面形精度直接影响遥感器性能。针对反射镜所处环境,组件一阶频率要求高,设计了一种反射镜背部中心单点挠性支撑结构,并利用拓扑优化技术对反射镜进行了轻量化设计。为了结构设计的合理性,采用有限元技术对反射镜组件进行了模态分析、结构强度分析、谐响应分析及热环境分析。分析结果表明:反射镜在重力作用下最大面形误差RMS=1.16nm;在±10℃温度载荷作用下最大面形误差RMS=11.46nm;反射镜组件一阶频率为574.19Hz。使挠性支撑结构具有良好的静、动态特性和热尺寸稳定性,满足反射镜面形要求。     

移动设备上Web服务的性能改进

随着移动设备在处理能力方面的迅速发展,内存和存储能力与更高的数据传输率结合在一起,如3G和4G无线技术方面的进步,目前在移动设备上托管Web服务是可行的.在本文中,我们提出了在移动设备上托管Web服务的轻量级框架.并在该框架中进一步对托管RESTful Web服务和基于SOAP的Web服务进行研究和比较分析.实验结果表明,RESTful Web服务对于在有限资源的移动设备上提供Web服务是低资源消耗和高效的.     

轻量级可移交CA的MANET网络认证体系

为解决移动自组网(Mobile Ad Hoc Network,MANET)网络信道开放、节点灵活多变且资源受限以及难以部署复杂认证机制的问题,结合轻量级CA思想,构造出一种适用于生存周期短、拓扑结构高度动态变化的MANET的认证体系结构即轻量级可移交认证中心(Lightweight and Shifted Certification Authority,LSCA)。LSCA结构简化了传统基于证书CA机制的公钥产生及验证的复杂性,无需证书管理;同时以移交CA角色的方式工作,不需预先配置节点及预知网络拓扑结构,使系统在不采用门限机制的情况下具备一定的容侵能力。性能分析及仿真实验表明:LSCA对DoS攻击表现出较强的健壮性,在通信、计算及存储代价方面均优于分布式CA及门限机制CA,适用于动态多变、生存周期较短的MANET网络应用。     

用不可能差分法分析12轮ESF算法

轻量级分组密码算法ESF是一种具有广义Feistel结构的32轮迭代型分组密码,轮函数具有SPN结构,分组长度为64比特,密钥长度为80比特。为了研究ESF算法抵抗不可能差分攻击的能力,基于一条8轮不可能差分路径,根据轮密钥之间的关系,通过向前增加2轮、向后增加2轮的方式,对12轮ESF算法进行了攻击。计算结果表明,攻击12轮ESF算法所需的数据复杂度为O(2⁵³),时间复杂度为O(260.43),由此说明12轮的ESF算法对不可能差分密码分析是不免疫的。     

PRINCE轻量级密码算法的差分故障分析

PRINCE密码算法是于ASIA CRYPT 2012提出的轻量级的加密算法,用于在物联网环境下保护RFID标签以及智能卡等设备的通信安全。提出并讨论了一种针对PRINCE算法的差分故障分析方法。该方法采用半字节故障模型,对PRINCEcore最后一轮进行了差分故障分析。实验结果表明,在PRINCEcore最后一轮导入半字节随机故障,4次故障注入可实现对PRINCE算法PRINCEcore部分的64位轮密钥的恢复。因此,未加防护措施的PRINCE加密系统将难以抵御差分故障分析手段。     

Studies on visual scene process system of aircraft assembly

According to the aircraft assembly worksite scene process, the hierarchical data model based on PPR (product, process, resources) was got to achieve the conversion in various types of aircraft assembly process data model. At the same time, the lightweight technology and human–computer interactive scene control technology were researched to obtain the lightweight of three-dimensional (3D) model and site scene and the human–computer real-time interaction between people and virtual environment through the computer keyboard and mouse, respectively. And then, the visual assembly sequence planning algorithm based on the parallel binary-coded was suggested to complete the aircraft assembly sequence planning. On the basis of techniques mentioned above, we developed the assembly worksite scene process system, which can be integrated with PDM system. At last, an application case of aircraft nose assembly was given, and the result was proved to be effective.