Secure multimedia
2 likes1,988 views
This document provides an overview of secure multimedia communications. It discusses various formats for images, audio, and video and considerations for securing multimedia content, including encryption, steganography, and watermarking. It also addresses challenges of multimedia security such as bandwidth requirements, throughput, packet loss, delay, and jitter. The document compares encryption, steganography, and watermarking and criteria for evaluating them such as visibility, robustness, and fragility.
Secure multimedia
- 1. Basic Training Editors: Deborah A. Frincke, Deborah.frincke@pnl.gov Richard Ford, rford@se.fit.edu Secure Multimedia Communications B ecause multimedia files can be shared with both representative of the value of multimedia and its cost. Im- increasing ease and efficiency, the need to en- age sizes can range from tens to thousands of kilobytes, and vid- sure their secure transmission and consump- eos can be orders of magnitude larger depending on factors such tion has become acute, particularly in light of as length, resolution, encoding method, and quality. Moreover, multimedia-specific exploits. This article explores a variety bandwidth requirements are com- pounded when multimedia files Liam M. of strategies and technologies that MP3, OGG, RM, and WAV); must be streamed at a consistent Mayron address the requirements and im- and rate. Streaming communications Harris plementation elements of secure • video (AVI, MOV, representa- are typically judged by the follow- Corporation multimedia communications. To tion as YCbCr, H.264, MPEG- ing characteristics:1 maintain acceptable performance, 4, RealVideo, QuickTime, and a security strategy must be aware WMV). • Throughput: the requisite mini- of multimedia-specific needs and mum data rate for the commu- not severely detriment metrics Image, video, and audio co- nication. such as jitter and frame rate. The decs are designed with different • Packet loss: the packets dropped wrong security techniques can intentions—for example, JPEG is while traversing the network. It burden the data with excessive en- a lossy encoding of image data (as isn’t typically optimal to retrans- coding that reduces its perceptual opposed to lossless). The process mit them due to the real-time quality and doesn’t achieve the in- of converting source image data nature of such communica- tended security objectives. into the JPEG format results in ir- tions, which impedes intelligible In this article, I provide an reversible data loss—it’s impossible transmission. overview of several elements of se- to reproduce the original image • Delay: the time between send- cure multimedia communications from a compressed JPEG image, as ing a packet from an originat- for network practitioners. The the compression process is a result ing host and receiving it at a wide variety of multimedia for- of removing data. The GIF format destination, or the round-trip mats available provides a range of employs a reduced color palette time for a transmission. Toler- performance trade-offs that must that’s suitable for synthetic, non- able elay varies depending on d be considered before selection and photographic images and doesn’t the environment and applica- then periodically revised. create the same type of artifacts as tion. For example, in voice over a JPEG. The SVG format allows IP, delay beyond 150 ms ren- Multimedia images to be described as vectors ders a call’s quality unaccept- Communications and scaled to different resolutions. able (www.cisco.com/en/US/ and Human Perception Audio and video formats often in- tech/tk869/tk769/technologies Multimedia data can be encoded clude features for compression and _wh it e _ paper 0 918 6 a 0 0 8 01 in a wide variety of formats, de- data protection and could be loss- b1a1e.shtml). pending on the media. Several less or lossy. • Jitter: inconsistent delivery times popular examples include (but are The most salient implication between packets. Although pack- not limited to) of multimedia communications ets can be transmitted at a con- compared to textual communi- sistent rate from the source host, • images (BMP, GIF, JPEG, PNG, cations is the higher bandwidth network conditions can intro- SVG, TIFF, and WebP); required. The phrase “a picture duce jitter, the results of which • audio (AAC, AIFF, AU, FLAC, is worth a thousand words” is can be as severe as packet loss.2 76 COPUBLISHED BY THE IEEE COMPUTER AND RELIABILITY SOCIETIES ■ 1540-7993/10/$26.00 © 2010 IEEE ■ NOVEMBER/DECEMBER 2010
- 2. Basic Training Employing a sufficiently large buf- cessful apparent motion is the ticated digital rights manage- fer for incoming data can reduce frame rate—to give the percep- ment (DRM) scheme, robust the effects of jitter. This buffer re- tion of smooth movement, most watermarks, and encryption that leases frames at a consistent rate, video frame rates are typically 24 eliminates any perceptually dis- but it introduces a slight additional frames per second or higher. For tinguishable elements. delay. Ultimately, jitter should be example, a video stream suffering • Resources. Some security meth- avoided whenever possible due to from jitter is at risk of not display- ods aren’t practical, given the the low tolerance people have for ing frames consistently. resources available (both compu- its effects.3 tational and economic). For ex- Various techniques share the Multimedia Security ample, encryption might not be common objective of efficiently Multimedia is used more and more feasible if bandwidth is limited. encoding multimedia in a way that throughout our daily life. The • Time. If the media is to be provides a tolerable and intelligi- volume of multimedia informa- broadcast live, encoding time ble experience for the recipient. A tion is massive when compared to might be a consideration. brief background on human visual traditional, textual data. The vari- perception can shed some light on ous codecs and protocols used for Encryption, steganography, and the design trade-offs of the dif- multimedia communications each watermarking are three examples ferent codecs. We can think of have their own potential vulner- of methods for securing multi- static images as two- imensional d abilities. The content being trans- media content. Encryption ob- arrays of discrete pixels, the nu- mitted could be extremely valuable scures the content from all except merical values that typically rep- or sensitive, or it might be freely the intended recipient. Steganog- resent an intensity or color at a available. Security schemes could raphy and watermarking per- certain location. An image array overburden networks or reduce the mit content to be broadcast and is analogous to the discrete recep- received quality. Approaches to en- viewed without requiring de- tors that interpret light that strikes suring the confidentiality, integri- cryption keys, but may include the retina4 (although receptors ty, and authenticity of multi edia m additional information (either aren’t as uniformly distributed depend on three key variables: overt or covert) for tracking the as pixels are). The human visual origin of the media or ensur- system perceives images in four • Value. The content’s value has a ing proper attribution. These stages from the original retinal direct impact on the investment techniques are basic approaches image produced by the receptors: expended to secure it. Valuable to securing multi edia content m the image-based stage (detection content might require a sophis- that must be understood before of image primitives and basic ge- ometry), the surface-based stage (stereoscopic, motion, texture, and shading factors), the object-based stage (3D and unseen surface con- sideration), and the category-based stage (pattern recognition and object identification). Lossy im- age codecs such as JPEG eliminate high-frequency image elements that are less essential to our under- standing of the image. The human visual system makes a distinction between static images and video due to apparent motion.4 Digital video is rendered as a series of consecutive frames: discrete, motionless images. Typi- cally, each frame is similar to the frame preceding it, with slight variations. Contemporary encod- ing techniques take advantage of this by keeping track of interframe disparities. The key factor of suc- www.computer.org/security 77
- 3. Basic Training investigating more sophisticated over, encryption could introduce content’s data size or maintain a DRM schemes. additional processing overhead similar amount of data. when decrypting. • Error tolerance. Many multi edia m Encryption For multimedia applications, codecs include mechanisms for Although encryption is often suf- practitioners should evaluate en- achieving resilient communica- ficient to ensure the confiden- cryption considering the follow- tions. Certain encryption schemes tiality of text communications, ing criteria: might disrupt this apability. c general-purpose encryption algo- rithms aren’t always appropriate • Content security. At the lowest Different multimedia encryp- in multimedia scenarios.5 En- level, a multimedia encryption tion algorithms satisfy these cri- cryption produces an apparently scheme might simply degrade teria to varying degrees and must random pattern of output bits, the content’s perceptual quality, be evaluated based on the specific and depending on when applied, or it could completely obscure needs of the target application and it might disrupt compression the content. multimedia formats employed. In schemes, resulting in very large • Format compliance. Format- practice, compression should al- file sizes (compression often de- compliant encryption preserves ways be performed before encryp- pends on the intrinsic, predictable the original media’s format and tion, not the reverse. Compression properties of the media’s content). permits encrypted content to reduces redundancies within con- The additional overhead incurred be communicated in the same tent, whereas good encryption by encryption can also adversely manner as unencrypted content. renders the content unintelligible, affect the criteria of multimedia • Data overhead. In some applica- leaving little that can be subse- communications—throughput, tions, encrypted content might quently compressed. Content that packet loss, delay, and jitter. More- need to preserve the original can be further compressed after encryption might not have been encrypted sufficiently. ADVERTISER INFORMATION • NOVEMBER/DECEMBER 2010 Steganography Steganography refers to a class of Advertising Personnel techniques for covertly embed- ding a payload within a cover Marian Anderson: Sr. Advertising Coordinator object.6 The cover object’s pur- Email: manderson@computer.org pose is to appear innocuous while Phone: +1 714 821 8380 | Fax: +1 714 821 4010 concealing the covert payload; the cover object can then be transmit- Sandy Brown: Sr. Business Development Mgr. ted across a public channel. Typi- Phone: +1 714 821 8380 | Fax: +1 714 821 4010 cally, the embedded payload is IEEE Computer Society much smaller than the cover ob- 10662 Los Vaqueros Circle ject. Steganography has particular Los Alamitos, CA 90720 application in multimedia content, USA where it can take advantage of www.computer.org data that has little perceptual sig- nificance. It allows the very exis- Advertising Sales Representatives tence of an encrypted message to be hidden. Western US/Pacific/Far East: Eric Kincaid A very simple example of im- Email: e.kincaid@computer.org age steganography relies on a Phone: +1 214 673 3742 protocol between the sender and Fax: +1 888 886 8599 intended recipient that flips the least significant bit of each pixel. Eastern US/Europe/Middle East: Ann David Schissler Email: a.schissler@computer.org, d.schissler@computer.org This change might look like noise Phone: +1 508 394 4026 to an unassuming viewer, but it Fax: +1 508 394 4926 can actually encode a text mes- sage, lower-resolution image, or other data. Significantly more sophisticated steganography tech- niques have been developed for 78 IEEE SECURITY PRIVACY
- 4. Basic Training images, audio, and video. For ex- • Fragile watermarks won’t be cryption Method for Digital ample, latent bandwidth in VoIP preserved if the content is modi- Videos,” Proc. Int’l Conf. Image applications provides unprece- fied, such as through transmis- Analysis and Recognition (ICIAR dented bandwidth for hiding mes- sion errors, lossy encoding, or 06), Springer, 2006, pp. 547–558. sages7 and could soon be exploited deliberate content manipulation. 6. N.F. Johnson and S. Jajodia, “Ex- for unintended purposes. Fragile watermarks ensure con- ploring Steganography: Seeing tent integrity. the Unseen,” Computer, vol. 31, Watermarking no. 2, 1998, pp. 26–34. Digital watermarks and steganog- Watermarks can form a com- 7. J. Lubacz, W. Mazurczyk, and raphy are extremely similar with ponent of a DRM system, as they K. Szczypiorski, “Vice over IP,” one notable difference: steganog- can validate the content’s owner as IEEE Spectrum, vol. 47, no. 2, raphy aims to hide the embedded well as the individual or organiza- 2010, pp. 42–47. payload’s presence, whereas water- tion with access to the content. 8. C.I. Podilchuk and E.J. Delp, marking exposes it. Watermarks “Digital Watermarking: Algo- should always be difficult to re- rithms and Applications,” IEEE move,8 ensuring proper attribution of content even if control over the distribution channel is lost. There S ecuring multimedia content presents challenges not pres- ent in network traffic such as web- 9. Signal Processing Magazine, vol. 18, no. 4, 2001, pp. 33–46. S.-J. Lee and S.-H. Jung, “A Sur- are three criteria for evaluating a pages, email, and other text. Not vey of Watermarking Techniques digital watermarking scheme:9 only is perceptual quality a con- Applied to Multimedia,” Proc. cern, but throughput, packet loss, Industrial Electronics Int’l Symp., • Visibility. A visible watermark delay, and jitter must also be con- IEEE Press, 2001, pp. 272–277. appears to the user when he or sidered. Encryption, steganogra- she records the content (for ex- phy, and watermarking are broad Liam M. Mayron is a member of the ample, the name of the com- classes of techniques that can be information and knowledge manage- pany that produced the media). used as part of a DRM solution. ment research group at Harris Corpo- An invisible one might be em- The specific methods employed ration. His research interests include bedded within the content in a depend on the application, avail- content-based image retrieval, cyber- manner that doesn’t reduce the able resources, and the value of security, biologically-inspired comput- media’s perceptual quality, much the content. Changing needs and ing, image processing, human and like steganography. new approaches (both for securing computer vision, and data mining. • Robustness. The watermark and exploiting content) necessitate Mayron has a PhD in computer engi- should be resilient to perturba- a periodic revision of employed neering from Florida Atlantic University. tions in the data. It shouldn’t be DRM schemes. Contact him at lmayron@harris.com. possible to remove the water- mark without distorting or de- References Selected CS articles and columns stroying the multimedia content. 1. W. Stallings, Data and Computer are also available for free at • Security. In some applications, Communications, 8th ed., Prentice http://ComputingNow.computer.org. it might be necessary to verify Hall, 2006. the authenticity of the authority 2. M. Claypool and J. Tanner, “The claiming the watermark. Effects of Jitter on the Perceptual Quality of Video,” Proc. ACM NEXT ISSUE We can assign watermarking Multimedia 99, ACM Press, 1999, algorithms to one of the following pp. 115–118. ENGINEERING classes: 3. R. Steinmetz, “Human Percep- SECURE tion of Jitter and Media Synchro- • Robust watermarks are intended nization Human Perception of SYSTEMS to be difficult to remove, even if Jitter and Media Synchroniza- there are specific changes made tion,” IEEE J. Selected Areas in to the content. Consequently, a Communications, vol. 14, no. 1, robust watermark can be used for 1996, pp. 61–72. access control and authentication. 4. S.E. Palmer, Vision Science: Pho- • Semi-fragile watermarks are resis- tons to Phenomenology, MIT Press, SUBSCRIBE AT tant to minor changes to content, 1999. WWW.COMPUTER.ORG/ but won’t be preserved if there are 5. D. Socek et al., “A Permutation- SECURITY significant modifications. Based Correlation-Preserving En- w ww.computer.org/security 79