多媒体指纹追踪：司法鉴定新领域(K. J. Liu)

需积分: 9 115 浏览量更新于2024-07-16 收藏 8.81MB PDF 举报

《叛徒追踪的多媒体指纹取证》（Multimedia Fingerprinting Forensics for Traitor Tracing）是一本专著，由K. J. Ray Liu、Wade Trappe、Z. Jane Wang、Min Wu和Hong Zhao共同编著，作为EURASIP信号处理与通信系列书籍的第四卷。这本书旨在向读者展示多媒体司法鉴定这个快速发展的领域中的关键见解和技术框架。多媒体指纹识别是本书的核心内容，它是一种利用多媒体数据的独特特征来追踪和识别潜在的信息泄露者或"叛徒"的技术。在信息技术时代，随着数据安全的重要性日益凸显，特别是对于版权保护、数字证据分析和反盗版等方面，多媒体指纹取证技术的应用变得至关重要。作者们详细探讨了如何通过创建和分析嵌入在音频、视频、图像等媒体中的不可见标记（指纹），来追踪那些未经授权分享或复制信息的行为。书中涵盖了多项关键技术，如水印技术（Watermarking）、数字签名（Digital Signatures）、频域分析（Frequency Domain Analysis）、时频分析（Time-Frequency Analysis）以及深度学习在多媒体指纹识别中的应用。这些技术被用来确保数据的完整性，同时还能追踪到数据的来源，有助于打击网络犯罪和知识产权侵权行为。此外，作者还介绍了相关的法律和伦理问题，因为多媒体指纹取证可能涉及到隐私权的冲突，以及如何在合法范围内进行追踪而不侵犯个人隐私。书中对这些问题进行了深入讨论，以期为法律实践者、信息安全专家以及研究人员提供一个全面的指导框架。本书不仅提供了理论基础，还包含了实际案例研究和未来发展趋势的预测，使得读者能够理解和掌握多媒体指纹取证在追踪叛徒过程中的应用策略。通过阅读这本书，读者可以提升对多媒体司法鉴定的理解，从而在这个领域中扮演更重要的角色，推动科技进步的同时维护社会秩序和知识产权的保护。

Preliminaries on data embedding

This chapter reviews the basics of robust data embedding. After a brief overview

on digital watermarking and data embedding technologies, we steer our a ttention

to a popular class of robust embedding techniques known as the spread-spectrum

embedding. The detailed formulation on the embedding and detection aspects of

the spread-spectrum technique establishes a foundation to unveil our technical

discussions on multimedia ﬁngerprinting in the subsequent chapters.

2.1. Content protection via digital watermarking

Multimedia content has both commercial and personal value that must be pro-

tected before one can share his/her work, or businesses can be founded to distrib-

ute and add value to their creations. Prior to digital multimedia content being put

onto the network for delivery, the data can be modiﬁed to help protect the intellec-

tual property of the content’s creators and service providers. Encryption and data

embedding are two complementary techniques for protecting multimedia content

that have diﬀerent goals. The primary goal behind encryption is conﬁdentiality

[29, 30, 31], that is, to provide access control so that only authorized users with

the correct decryption keys can access the content. The protection provided by

encryption terminates after decryption. Complementing this functionality, data

embedding or digital watermarking associates a set of secondary data with the

host media in a seamless way [17, 18]. The term “digital watermark” comes from

an analogy to its analog counterpart: as an art of paper making, paper watermarks

usually indicate the origin and the ownership, and/or establish the integrity and

prevent counterfeiting. Similarly, digital watermarking has been considered in s ev-

eral real-world applications related to multimedia content protection and security.

These include copy prevention for DVD and digital music, the assertion of owner-

ship, the ﬁngerprinting and tracing of content recipients, and the authentication

of the content. While the protection provided by watermarks is usually passive, the

embedded watermarks can travel with the host media and assume their protection

function even after decryption. This capability of associating additional data with

8 Preliminaries on data embedding

the digital multimedia content in a seamless way is so unique that few conventional

protection tools for generic data, such as cryptographic encryption, can oﬀer it.

Digital watermarking technologies, or more generally, the data embedding

technologies, can be applied for a number of applications, each of w hich has its

own design requirements in terms of the imperceptibility, the robustness, and the

embedding payload measured in terms of how many bits are embedded. As the

embedding process more or less changes the original multimedia signal (known as

the host signal), it is important in a number of applications to conﬁne the changes

to below the perceivable levels of human in order to preserve the commercial and

aesthetic values of the host multimedia signal. In the following, we brieﬂy review

a few major applications and approaches of data embedding for multimedia.

2.1.1. Major applications and design requirements

Ownership protection. In this application, a watermark signal is secretly selected

by the copyright holder to represent his/her ownership, and is imperceptibly em-

bedded in the multimedia source. As pirates are highly motivated to remove the

sist intentional attacks so that the owner can demonstrate the presence of this wa-

termark in case of dispute to verify his/her ownership. The detection should have

as little ambiguity and false alarm as possible. In most scenarios, the total number

of bits that can be embedded and extrac ted reliably does not have to be high.

Authentication or tampering detection. In this application, we embed a set of sec-

ondary data in the multimedia source beforehand, and later use it to determine

whether the host media is tampered or not. The robustness against removing the

watermark or making it undetectable is not a major concern as this clearly sig-

nals the occurrence of tampering and there is no strong incentive to do so from

an attacker’s point of view. The main threat is the forging of a valid authentica-

tion watermark in an u nauthorized or tampered multimedia signal, which must

be prevented. In many practical applications, it is also desirable to locate the tam-

pered regions and distinguish some noncontent changes (such as those incurred

by moderate lossy compression) from other changes (such as content tampering).

In general, the embedding payload should be suﬃciently high to accommodate

these needs. The detection must be performed without using the original unwa-

termarked copy, because either this original is unavailable or its integrity has not

been established yet. This kind of detection is known as noncoherent detection or

blind detection.

Digital ﬁngerprinting. As we have explained in the introductory chapter, a water-

mark in this application serves as a ﬁngerprint to help trace the originator or recip-

ients of a particular copy of multimedia content. The robustness against removal

and the ability to convey a nontrivial number of bits are necessary requirements. In

addition, digital ﬁngerprinting techniques should also be robust against collusion

when users having access to the same host image embedded with diﬀerent ﬁnger-

print IDs get together and try to remove the ﬁngerprints through such operations

Content protection via digital watermarking 9

as averaging. To assure the reliable tracing of true traitors and avoid framing in-

nocents, we must carefully design how to construct, embed, and detect ﬁngerprint

signals. This suggests that there are more considerations than those that the robust

embedding technologies oﬀered in the existing digital watermarking literature. We

will devote our a ttention to these issues in the rest of this book.

Copy control and access control. The embedded watermark in this application rep-

resents certain copy control or access control policy. A watermark detector is of-

ten integrated in a recording or playback system. Upon detection, the policy is

enforced by directing certain hardware or software actions such as enabling or

disabling a recording module. The robustness against removal, the ability of blind

detection, and the capability of conveying a nontrivial number of bits are required.

Two examples from the past standardization eﬀorts are the DVD copy control [32]

and the Secure Digital Music Initiatives (SDMI) a ctivities [9].

Annotation. The embedded watermark in this application is expected to convey

as many bits as possible without the use of original unmarked copy in detection.

While the robustness against intentional attack is not required, many scenarios

may prefer a certain degree of robustness against common processing such as lossy

compression. More generally, data embedding is a tool to convey side information

while retaining the original appearance. This property has been found useful in

multimedia communications [33, 34, 35] to achieve additional functionalities, or

improve security and performance.

2.1.2. Basic embedding approaches

A typical data hiding framework is illustrated in Figure 2.1. Starting with an orig-

inal digital media ( I

), which is also known as the host media or cover work, the

embedding module inserts in it a set of secondary data (b

), which is referred to

as embedded data or watermark, to obtain the marked media (I

). The insertion

or embedding is done such that I

is perceptually identical to I

.Thediﬀerence

between I

and I

is the distortion introduced by the embedding process. In most

cases, the embedded data is a collection of bits, which may come from an encoded

character string, from a pattern, or from some executable agents, depending on

the application.

The embedded data b

will be extracted from the marked media I

by a de-

tector, often after I

has gone through various processing and attacks. The input

to the detector is referred to as test media (I

), and the extracted data from I

denoted by



b.Thediﬀerence between I

and I

is called noise. In such applications

as ownership protection, ﬁngerprinting, and access control, accurate decoding of

hidden data from distorted test media is preferred. In other applications such as

authentication and annotation, robustness is not critical.

The embedding of one bit in host media is basic to every data hiding system.

Almost all embedding approaches belong to one of two general types. Below we

examine them in more detail.

Robust additive spread-spectrum embedding 11

subset that maps to b, so that the relationship of b

= g(I

) is deterministically en-

forced. To minimize perceptual distortion, I

should be as close to I

as possible.

A simple example is odd-e ven embedding, whereby a closest even number is used

as I

to embed a “0” and a closest odd number is used to embed a “1.” The em-

bedded bit is extracted simply by checking the odd-even parity, w hich does not

require the knowledge of original I

. There can be other constraints imposed on

for robustness considerations. For example, the enforcement can be done in a

quantized domain with a step size Q [18, 37]. The odd-even enforcement can also

be extended to higher dimensions involving features computed from a set of host

components.

Under blind detection where the host signal is not available at the detector

and becomes a major source of noise for Type-I embedding, the number of bits

that can be reliably embedded by Type-II is much higher than that of Type-I

when the noise is not strong [18, 38, 39]. On the other hand, Type-I is com-

monly used for robust embedding with strong noise (especially when noise be-

comes much stronger than the watermark) as well as for nonblind detection. Mo-

tivated by Costa’s information-theoretical result [40], distortion compensation

has been proposed to be incorporated into quantization-based enforcement em-

bedding [39, 41, 42], where the enforcement is combined linearly with the host

signal to form a watermarked signal. The optimal scaling factor is a func tion of

WNR and will increase the number of bits that can be embedded. This distortion-

compensated embedding can be viewed a s a combination of Type-I and Type-II

embedding.

2.2. Robust additive spread-spectrum embedding

With a big picture in mind, we now zoom in to the problem of robust embedding,

which serves as an important building block for embedding ﬁngerprints into the

multimedia content.

Fingerprinting multimedia requires the use of robust data embedding meth-

ods that are capable of withstanding attacks that adversaries might employ to re-

move the ﬁngerprint. Although there are many techniques that have been pro-

posed for embedding information in multimedia signals [17], in the sequel we

will use the spread-spectrum additive embedding technique for illustrating the

embedding of ﬁngerprint signals into multimedia. Spread-spectr um embedding

has proven robust against a number of signal processing operations (such as lossy

compression and ﬁltering) and attacks [23, 24]. With appropriately chosen fea-

tures and additional alignment procedures, the spread-spectrum watermark can

survive moderate geomet ric distortions, such as rotation, scale, shift, and cropping

[43, 44]. Further, information-theoretic studies suggest that it is nearly capacity

optimal when the original host s ignal is available in detection [38, 39]. The combi-

nation of robustness and c apacity makes spread-spectrum embedding a promising

technique for protecting multimedia. In addition, as we will see in later chapters,

its capability of putting multiple marks in overlapped regions also limits the eﬀec-

tive strategies mountable by colluders in ﬁngerprinting applications.

剩余259页未读，继续阅读

weixin_38744207

粉丝: 344
资源: 2万+

多媒体指纹追踪：司法鉴定新领域(K. J. Liu)

halsted.press.h.264.and.mpeg-4.video.compression.video.coding.for.next.generation.multimedia.ebook-lib.rar

test config.qtmultimedia_multimedia.libraries.libresourceqt5 failed

https://ir.ozone.ru/s3/multimedia-7/wc350/6494414143.jpg 我想把这个地址转成图片显示在EXCEL 里面

MSE被用于图像融合评价函数的文献

人脸老化和退龄预测参考文献

关于多媒体素材管理系统的设计与实现的参考文献

要写基于pytorch的ocr文字识别的开题报告，参考文献有哪些

基于特征匹配的彩色图像拼接方法研究参考文献

"QtMultimedia" version 5.15 is not installed

与“基于HarmonyOS的视频播放器的设计与实现”相关的英文文献

最新资源