假指纹与活体指纹检测

目录

[1. 假指纹简介](#1. 假指纹简介)

[2. 假指纹制作流程](#2. 假指纹制作流程)

[3. 活体指纹检测](#3. 活体指纹检测)

[4. 活体指纹检测竞赛](#4. 活体指纹检测竞赛)


1. 假指纹简介

随着科学技术的发展,指纹技术以各种各样的形式进入了我们的生活。在大多数情况下,指纹识别应用于移动设备和桌面设备解决方案,以提供安全方便的认证。

然而,如今的指纹传感器很容易被虚假指纹欺骗,虚假指纹的欺诈风险对移动支付等敏感应用程序构成威胁。

虚假指纹有多种制造方法。据报道,可以使用日常廉价的材料(如明胶、硅胶、橡皮泥等)制造逼真的伪指纹,足以骗过指纹识别系统(见下图)。例如,2013年3月,一名巴西医生因使用硅胶制成的伪指纹欺骗圣保罗一家医院的指纹考勤系统而被捕(BBC新闻,2013)。2013年9月,苹果发布内置Touch ID指纹技术的iPhone 5S后不久,德国的Chaos计算机俱乐部(CCC,2013)根据注册用户的高分辨率指纹照片用木胶制作了伪指纹,成功欺骗了Touch ID系统。2016年7月,密歇根州立大学的研究人员使用2D打印指纹解锁了一部智能手机,以帮助警方处理凶杀案(Korkzan,2016)。2018年3月,印度拉贾斯坦邦的一个团伙通过蜡模中注入胶水制作伪指纹欺骗警方的指纹考勤系统,因此而被捕(Vidyut,2018)。很可能还有大量的虚假指纹攻击未被发现,因此没有报告。

虚假指纹可以由多种常见材料制作

2.指纹制作流程

假指纹制作可以在目标用户合作时进行。

3. 活体指纹检测

识别假指纹的方法主要有两种模式,即硬件模式和软件模式。其中,硬件模式需要设计专门的硬件设计,并与生物特征识别传感器进行整合,但是设备更新难度较大;软件模式则是对指纹传感器获取的图像进行分析,对真假指纹进行对比,通过大量的学习,识别真假指纹。由于软件模式更新更方便,因而得到更广泛的客户认可。

通过软件识别假指纹是通过对提取的指纹特征数据进行分析,判断真假指纹。由于假指纹存在一定程度的失真,因而会导致特征点丢失,另外,假指纹会加入一定的杂讯(例如橡皮泥做的假指纹存在气泡),因而可以通过大量机器学习及人工智能形成数据基础来识别假指纹。

由于可能的伪造材料很多、不同指纹的传感器的差异很大,为提升活体检测技术的泛化能力,规模更大、种类更多的假指纹数据库有待开发。

4. 活体指纹检测竞赛

LivDet(LivDet - Liveness Detection Competitions)赞助的"活体检测竞赛"(Liveness Detection Competitions),是一项两年一度的国际竞赛,向学术界和工业界开放,旨在评估和报告指纹呈现攻击检测方面的进展。LivDet还将基准测试的数据开放给设备制造商,帮助工程师能够达到理想的"误拒绝"和成功率设计目标。

LivDet 2009

|-------|------------|--------------|----------|--------|------------|
| | Sensor | type | size | ID | Images |
| Train | Biometrika | live | 312x372 | 13 | 520 |
| Train | Biometrika | Silicone | 312x372 | 13 | 520 |
| Train | CrossMatch | live | 640x480 | 125 | 1000 |
| Train | CrossMatch | Gelatin(明胶) | 640x480 | 33 | 344 |
| Train | CrossMatch | PlayDoh(泥胶) | 640x480 | 30 | 346 |
| Train | CrossMatch | Silicone(硅胶) | 640x480 | 15 | 310 |
| Train | Identix | live | 720x720 | 69 | 750 |
| Train | Identix | Gelatin | 720x720 | 37 | 250 |
| Train | Identix | PlayDoh | 720x720 | 32 | 250 |
| Train | Identix | Silicone | 720x720 | 15 | 250 |
| Test | Biometrika | live | 312x372 | 37 | 1473 |
| Test | Biometrika | Silicone | 312x372 | 37 | 1480 |
| Test | CrossMatch | live | 312x372 | 377 | 3000 |
| Test | CrossMatch | Gelatin(明胶) | 312x372 | 106 | 1036 |
| Test | CrossMatch | PlayDoh(泥胶) | 312x372 | 102 | 1034 |
| Test | CrossMatch | Silicone(硅胶) | 312x372 | 56 | 932 |
| Test | Identix | live | 720x720 | 250 | 2250 |
| Test | Identix | Gelatin | 720x720 | 105 | 750 |
| Test | Identix | PlayDoh | 720x720 | 99 | 750 |
| Test | Identix | Silicone | 720x720 | 60 | 750 |

LivDet 2011

|-------|------------|---------------|----------|--------|------------|
| | Sensor | type | size | ID | Images |
| Train | Biometrika | live | 312x372 | 200 | 1000 |
| Train | Biometrika | EcoFlex(降解塑料) | 312x372 | 20 | 200 |
| Train | Biometrika | Gelatin(明胶) | 312x372 | 20 | 200 |
| Train | Biometrika | Latex(胶乳) | 312x372 | 20 | 200 |
| Train | Biometrika | Silgum | 312x372 | 20 | 200 |
| Train | Biometrika | WoodGlue(木胶) | 312x372 | 20 | 200 |
| Train | Italdata | live | 640x480 | 200 | 1000 |
| Train | Italdata | EcoFlex(降解塑料) | 640x480 | 20 | 200 |
| Train | Italdata | Gelatin(明胶) | 640x480 | 20 | 200 |
| Train | Italdata | Latex(胶乳) | 640x480 | 20 | 200 |
| Train | Italdata | Silgum | 640x480 | 20 | 200 |
| Train | Italdata | WoodGlue(木胶) | 640x480 | 20 | 200 |
| Train | Digital | live | 355x391 | 84 | 1004 |
| Train | Digital | Gelatin(明胶) | 355x391 | 26 | 200 |
| Train | Digital | Latex(胶乳) | 355x391 | 22 | 200 |
| Train | Digital | PlayDoh(泥胶) | 355x391 | 20 | 200 |
| Train | Digital | Silicone(硅胶) | 355x391 | 24 | 200 |
| Train | Digital | Wood Glue(木胶) | 355x391 | 26 | 200 |
| Train | Sagem | live | 352x384 | 58 | 1008 |
| Train | Sagem | Gelatin(明胶) | 352x384 | 38 | 200 |
| Train | Sagem | Latex(胶乳) | 352x384 | 20 | 201 |
| Train | Sagem | PlayDoh(泥胶) | 352x384 | 42 | 200 |
| Train | Sagem | Silicone(硅胶) | 352x384 | 28 | 200 |
| Train | Sagem | Wood Glue(木胶) | 352x384 | 22 | 207 |
| test | Biometrika | live | 312x372 | 200 | 1000 |
| test | Biometrika | EcoFlex(降解塑料) | 312x372 | 20 | 200 |
| test | Biometrika | Gelatin(明胶) | 312x372 | 20 | 200 |
| test | Biometrika | Latex(胶乳) | 312x372 | 20 | 200 |
| test | Biometrika | Silgum | 312x372 | 20 | 200 |
| test | Biometrika | WoodGlue(木胶) | 312x372 | 20 | 200 |
| test | Italdata | live | 640x480 | 200 | 1000 |
| test | Italdata | EcoFlex(降解塑料) | 640x480 | 20 | 200 |
| test | Italdata | Gelatin(明胶) | 640x480 | 20 | 200 |
| test | Italdata | Latex(胶乳) | 640x480 | 20 | 200 |
| test | Italdata | Silgum | 640x480 | 20 | 200 |
| test | Italdata | WoodGlue(木胶) | 640x480 | 20 | 200 |
| test | Digital | live | 355x391 | 104 | 1000 |
| test | Digital | Gelatin(明胶) | 355x391 | 28 | 200 |
| test | Digital | Latex(胶乳) | 355x391 | 20 | 200 |
| test | Digital | PlayDoh(泥胶) | 355x391 | 20 | 200 |
| test | Digital | Silicone(硅胶) | 355x391 | 34 | 200 |
| test | Digital | Wood Glue(木胶) | 355x391 | 24 | 200 |
| test | Sagem | live | 352x384 | 40 | 1000 |
| test | Sagem | Gelatin(明胶) | 352x384 | 46 | 225 |
| test | Sagem | Latex(胶乳) | 352x384 | 20 | 204 |
| test | Sagem | PlayDoh(泥胶) | 352x384 | 42 | 205 |
| test | Sagem | Silicone(硅胶) | 352x384 | 56 | 200 |
| test | Sagem | Wood Glue(木胶) | 352x384 | 20 | 202 |

LivDet 2013

|-------|------------|---------------|----------|--------|------------|
| | Sensor | type | size | ID | Images |
| train | Biometrika | live | 312x372 | 200 | 1000 |
| train | Biometrika | EcoFlex(降解塑料) | 312x372 | 20 | 200 |
| train | Biometrika | Gelatin(明胶) | 312x372 | 20 | 200 |
| train | Biometrika | Latex(胶乳) | 312x372 | 20 | 200 |
| train | Biometrika | Silgum | 312x372 | 20 | 200 |
| train | Biometrika | WoodGlue(木胶) | 312x372 | 20 | 200 |
| train | CrossMatch | live | 800x750 | 484 | 1250 |
| train | CrossMatch | BodyDouble | 800x750 | 125 | 250 |
| train | CrossMatch | Latex(胶乳) | 800x750 | 125 | 250 |
| train | CrossMatch | Playdoh(泥胶) | 800x750 | 125 | 250 |
| train | CrossMatch | WoodGlue(木胶) | 800x750 | 125 | 250 |
| train | Italdata | live | 640x480 | 200 | 1000 |
| train | Italdata | Ecoflex(降解塑料) | 640x480 | 20 | 200 |
| train | Italdata | Gelatin(明胶) | 640x480 | 20 | 200 |
| train | Italdata | Latex(胶乳) | 640x480 | 20 | 200 |
| train | Italdata | Modasil | 640x480 | 20 | 200 |
| train | Italdata | WoodGlue (木胶) | 640x480 | 20 | 200 |
| train | Swipe | live | 208x1500 | 247 | 1221 |
| train | Swipe | BodyDouble | 208x1500 | 125 | 250 |
| train | Swipe | Latex(胶乳) | 208x1500 | 125 | 250 |
| train | Swipe | Playdoh(泥胶) | 208x1500 | 119 | 233 |
| train | Swipe | Wood Glue(木胶) | 208x1500 | 124 | 246 |
| test | Biometrika | live | 312x372 | 200 | 1000 |
| test | Biometrika | EcoFlex(降解塑料) | 312x372 | 20 | 200 |
| test | Biometrika | Gelatin(明胶) | 312x372 | 20 | 200 |
| test | Biometrika | Latex(胶乳) | 312x372 | 20 | 200 |
| test | Biometrika | Silgum | 312x372 | 20 | 200 |
| test | Biometrika | WoodGlue(木胶) | 312x372 | 20 | 200 |
| test | CrossMatch | live | 800x750 | 430 | 1250 |
| test | CrossMatch | BodyDouble | 800x750 | 63 | 250 |
| test | CrossMatch | Latex(胶乳) | 800x750 | 65 | 250 |
| test | CrossMatch | Playdoh(泥胶) | 800x750 | 63 | 250 |
| test | CrossMatch | WoodGlue(木胶) | 800x750 | 63 | 250 |
| test | Italdata | live | 640x480 | 200 | 1000 |
| test | Italdata | Ecoflex(降解塑料) | 640x480 | 20 | 200 |
| test | Italdata | Gelatin(明胶) | 640x480 | 20 | 200 |
| test | Italdata | Latex(胶乳) | 640x480 | 20 | 200 |
| test | Italdata | Modasil | 640x480 | 20 | 200 |
| test | Italdata | WoodGlue (木胶) | 640x480 | 20 | 200 |
| test | Swipe | live | 208x1500 | 235 | 1153 |
| test | Swipe | BodyDouble | 208x1500 | 82 | 250 |
| test | Swipe | Latex(胶乳) | 208x1500 | 91 | 250 |
| test | Swipe | Playdoh(泥胶) | 208x1500 | 87 | 250 |
| test | Swipe | Wood Glue(木胶) | 208x1500 | 75 | 250 |

LivDet 2015

|-------|-----------------|----------------|-----------|--------|------------|
| | Sensor | type | size | ID | Images |
| train | CrossMatch | Live | 800x750 | 500 | 1510 |
| train | CrossMatch | Body Double | 800x750 | 165 | 494 |
| train | CrossMatch | Ecoflex | 800x750 | 168 | 498 |
| train | CrossMatch | Playdoh | 800x750 | 166 | 481 |
| train | Digital_Persona | Live | 252x324 | 100 | 1000 |
| train | Digital_Persona | Ecoflex 00-50 | 252x324 | 80 | 250 |
| train | Digital_Persona | Gelatine | 252x324 | 80 | 250 |
| train | Digital_Persona | Latex | 252x324 | 80 | 250 |
| train | Digital_Persona | WoodGlue | 252x324 | 80 | 250 |
| train | GreenBit | Live | 500x500 | 100 | 1000 |
| train | GreenBit | Ecoflex 00-50 | 500x500 | 80 | 250 |
| train | GreenBit | Gelatine | 500x500 | 80 | 250 |
| train | GreenBit | Latex | 500x500 | 80 | 250 |
| train | GreenBit | WoodGlue | 500x500 | 80 | 250 |
| train | Hi_Scan | Live | 1000x1000 | 100 | 1000 |
| train | Hi_Scan | Ecoflex 00-50 | 1000x1000 | 80 | 250 |
| train | Hi_Scan | Gelatine | 1000x1000 | 80 | 250 |
| train | Hi_Scan | Latex | 1000x1000 | 80 | 250 |
| train | Hi_Scan | WoodGlue | 1000x1000 | 80 | 250 |
| train | Time_Series | Live | 800x750 | 500 | 4440 |
| train | Time_Series | Body Double | 800x750 | 165 | 1481 |
| train | Time_Series | Ecoflex | 800x750 | 170 | 1529 |
| train | Time_Series | Playdoh | 800x750 | 165 | 1485 |
| test | CrossMatch | Live | 800x750 | 500 | 1500 |
| test | CrossMatch | Body Double | 800x750 | 100 | 300 |
| test | CrossMatch | Ecoflex | 800x750 | 92 | 270 |
| test | CrossMatch | Gelatin | 800x750 | 100 | 300 |
| test | CrossMatch | OOMOO | 800x750 | 100 | 297 |
| test | CrossMatch | Playdoh | 800x750 | 95 | 281 |
| test | Digital_Persona | Live | 252x324 | 100 | 1000 |
| test | Digital_Persona | Ecoflex 00-50 | 252x324 | 80 | 250 |
| test | Digital_Persona | Gelatine | 252x324 | 80 | 250 |
| test | Digital_Persona | Latex | 252x324 | 80 | 250 |
| test | Digital_Persona | Liquid Ecoflex | 252x324 | 80 | 250 |
| test | Digital_Persona | RTV | 252x324 | 80 | 250 |
| test | Digital_Persona | WoodGlue | 252x324 | 80 | 250 |
| test | GreenBit | Live | 500x500 | 100 | 1000 |
| test | GreenBit | Ecoflex 00-50 | 500x500 | 80 | 250 |
| test | GreenBit | Gelatine | 500x500 | 80 | 250 |
| test | GreenBit | Latex | 500x500 | 80 | 250 |
| test | GreenBit | Liquid Ecoflex | 500x500 | 80 | 250 |
| test | GreenBit | RTV | 500x500 | 80 | 250 |
| test | GreenBit | WoodGlue | 500x500 | 80 | 250 |
| test | Hi_Scan | Live | 1000x1000 | 100 | 1000 |
| test | Hi_Scan | Ecoflex 00-50 | 1000x1000 | 80 | 250 |
| test | Hi_Scan | Gelatine | 1000x1000 | 80 | 250 |
| test | Hi_Scan | Latex | 1000x1000 | 80 | 250 |
| test | Hi_Scan | Liquid Ecoflex | 1000x1000 | 80 | 250 |
| test | Hi_Scan | RTV | 1000x1000 | 80 | 250 |
| test | Hi_Scan | WoodGlue | 1000x1000 | 80 | 250 |

相关推荐
李加号pluuuus2 小时前
【扩散模型】LCM LoRA:一个通用的Stable Diffusion加速模块
人工智能·stable diffusion
Alkali!3 小时前
2-5 softmax 回归的简洁实现
人工智能·数据挖掘·回归
哥廷根数学学派4 小时前
基于Maximin的异常检测方法(MATLAB)
开发语言·人工智能·深度学习·机器学习
xrgs_shz4 小时前
人工智能、机器学习、神经网络、深度学习和卷积神经网络的概念和关系
人工智能·深度学习·神经网络·机器学习·卷积神经网络
zzyincsdn5 小时前
从FasterTransformer源码解读开始了解大模型(2.1)代码通读03
人工智能·llm·fastertransform
dc爱傲雪和技术6 小时前
相机参数与图像处理技术解析
图像处理·人工智能·数码相机
QQ21696368816 小时前
Hi6602 恒压恒流SSR电源方案
人工智能·硬件架构·硬件工程·信息与通信·智能硬件
wodertianna6 小时前
机器学习原理之 -- 神经网络:由来及原理详解
人工智能·神经网络·机器学习
续亮~6 小时前
9、程序化创意
前端·javascript·人工智能
muren7 小时前
昇思MindSpore学习笔记2-01 LLM原理和实践 --基于 MindSpore 实现 BERT 对话情绪识别
笔记·深度学习·学习