本书从物联网技术发展现状、体系架构及演进趋势入手，设计物联网安全架构；在对物联网典型安全事件进行回顾的基础上，梳理物联网安全问题分类，提炼安全威胁和安全需求，提出物联网安全体系框架，引出物联网身份安全的重要性；深入介绍物联网安全认证技术，从传统身份认证机制、物联网身份认证方法入手，对基于生物特征的物联网身份认证方法和基于深度学习的声纹识别技术进行详细描述；介绍了生物特征识别技术中能够实用化、商业化的深度学习算法，并对典型的深度学习框架和平台进行了分析。

吴巍，中国电子科技集团有限公司第五十四研究所研究员，副总工程师，博士生导师，通信网信息传输与分发技术国家重点实验室主任，中国电子科技集团公司首席科学家。长期从事通信网技术研究与装备研制工作，研究领域涉及通信网总体、组网与信息分发、路由交换、通信网络安全技术等。先后主持完成了多项国家重大通信与信息系统科研与工程项目，获国家科技进步二等奖1项，省部级一等奖5项、二等奖6项。发表论文50余篇，获发明专利授权19项，出版专著3部。1993年获国防科工委《光华科技基金会》二等奖，1996年享受国务院特殊津，2002年获国防科工委"有突出贡献中青年专家”称号。

第1 章 物联网技术基础···········································································.001
1.1 物联网发展现状··········································································.001
1.1.1 美国·················································································.001
1.1.2 欧盟·················································································.002
1.1.3 日本·················································································.002
1.1.4 韩国·················································································.003
1.1.5 中国·················································································.003
1.2 物联网体系架构··········································································.004
1.2.1 感知层·············································································.005
1.2.2 网络层·············································································.005
1.2.3 应用层·············································································.006
1.3 物联网信息的三大特性·······························································.007
1.3.1 高敏感性··········································································.007
1.3.2 高可靠性··········································································.007
1.3.3 高关联性··········································································.007
1.4 物联网体系架构的发展·······························································.008
1.5 小结·······························································································011
第2 章 物联网安全架构··········································································.012
2.1 引言·····························································································.012
2.2 物联网典型安全事件···································································.012
2.2.1 事件回顾··········································································.012
2.2.2 事件分析··········································································.016
2.3 物联网安全问题分类···································································.017
2.3.1 互联网引入的安全问题···················································.017
2.3.2 物联网场景下的互联网“安全”问题····························.017
2.3.3 物联网引入的安全问题···················································.017
2.3.4 物联网自身的安全问题···················································.018
2.4 物联网安全威胁分析···································································.018
2.4.1 感知层安全威胁······························································.019
2.4.2 网络层安全威胁······························································.020
2.4.3 应用层安全威胁······························································.022
2.5 物联网安全需求分析···································································.023
2.5.1 纵横联动的一体化安全保障支撑····································.024
2.5.2 感知层传感器设备的身份鉴别与数据防护····················.026
2.5.3 网络层异构网络规模化安全互联与全网统一监管·········.027
2.5.4 应用层数据多域安全共享···············································.028
2.6 物联网安全体系框架···································································.029
2.6.1 技术体系··········································································.029
2.6.2 物联网系统的安全信息流···············································.036
2.7 小结·····························································································.037
第3 章 物联网安全认证技术···································································.039
3.1 引言·····························································································.039
3.2 身份认证方式··············································································.040
3.2.1 基于秘密信息的认证方式···············································.040
3.2.2 基于信物的认证方式·······················································.041
3.2.3 基于密钥的认证方式·······················································.041
3.2.4 基于生物特征的认证方式···············································.042
3.3 物联网身份认证的特点·······························································.043
3.3.1 轻量级·············································································.043
3.3.2 非对称·············································································.043
3.3.3 复杂性·············································································.043
3.4 几种物联网身份认证方式···························································.044
3.4.1 基于RFID 的物联网身份认证方式·································.044
3.4.2 基于传感网络的物联网认证方式····································.044
3.4.3 基于蓝牙的感知网络认证方式·······································.045
3.4.4 基于生物特征识别的认证方式·······································.045
3.5 基于生物特征的物联网身份认证方法········································.047
3.5.1 生物特征身份认证流程···················································.047
3.5.2 指纹识别··········································································.049
3.5.3 人脸识别··········································································.049
3.5.4 虹膜识别··········································································.051
3.5.5 指静脉识别······································································.051
3.5.6 声纹识别··········································································.052
3.6 基于深度学习的声纹识别技术····················································.053
3.6.1 概述·················································································.053
3.6.2 声纹识别的工作原理·······················································.054
3.6.3 声纹识别的流程······························································.055
3.6.4 声纹识别技术的三次突破···············································.059
3.6.5 基于深度学习的典型声纹识别算法································.060
3.6.6 声纹识别应用趋势··························································.063
3.7 小结·····························································································.066
本章参考文献······················································································.066
第4 章 卷积神经网络技术······································································.068
4.1 卷积运算······················································································.069
4.2 动机·····························································································.072
4.3 池化·····························································································.077
4.4 将卷积与池化作为一个无限强的先验········································.082
4.5 基本卷积函数的变体···································································.083
4.6 结构化输出··················································································.093
4.7 数据类型······················································································.094
4.8 高效的卷积算法··········································································.095
4.9 随机或无监督的特征···································································.096
4.10 小结···························································································.097
本章参考文献······················································································.098
第5 章 序列建模：循环和递归网络·······················································.101
5.1 展开计算图··················································································.102
5.2 RNN·····························································································.105
5.2.1 Teacher Forcing 和输出循环网络····································.109
5.2.2 计算RNN 的梯度······························································111
5.2.3 作为有向图模型的循环网络·············································113
5.2.4 基于上下文的RNN 序列建模···········································117
5.3 双向RNN ······················································································119
5.4 基于编码-解码的序列到序列架构···············································.121
5.5 深度RNN ····················································································.123
5.6 递归神经网络··············································································.124
5.7 长期依赖的挑战··········································································.126
5.8 回声状态网络··············································································.128
5.9 渗漏单元和其他多时间尺度的策略············································.130
5.9.1 时间维度的跳跃连接·······················································.130
5.9.2 渗漏单元和一系列时间尺度···········································.131
5.9.3 删除连接··········································································.131
5.10 长短期记忆和其他门控RNN ····················································.132
5.10.1 长短期记忆····································································.133
5.10.2 其他门控RNN·······························································.135
5.11 优化长期依赖············································································.136
5.11.1 截断梯度········································································.136
5.11.2 引导信息流的正则化·····················································.138
5.12 外显记忆····················································································.139
5.13 小结···························································································.142
本章参考文献······················································································.142
第6 章 深度学习框架和平台的分析与对比············································.148
6.1 概述·····························································································.148
6.2 深度学习框架··············································································.151
6.2.1 TensorFlow·······································································.152
6.2.2 Caffe ················································································.156
6.2.3 PyTorch ············································································.162
6.2.4 CNTK···············································································.164
6.2.5 MXNet ·············································································.166
6.3 深度学习框架的分析与对比·······················································.169
6.3.1 总体分析··········································································.169
6.3.2 深度学习框架的对比·······················································.170
6.3.3 深度学习框架对硬件的利用情况····································.178
6.4 深度学习平台··············································································.180
6.4.1 华为深度学习服务DLS ··················································.180
6.4.2 阿里深度学习开发平台X-DeepLearning ························.184
6.4.3 百度深度学习开发平台PAddle·······································.189
6.4.4 几种平台的对比······························································.191
6.5 小结·····························································································.192