Research Article A Novel Steganography Technique for SDTV ...downloads.hindawi.com/journals/ijdmb/2016/6950592.pdf · Research Article A Novel Steganography Technique for SDTV-H.264/AVC

Research ArticleA Novel Steganography Technique forSDTV-H264AVC Encoded Video

Christian Di Laura Diego Pajuelo and Guillermo Kemper

School of Electrical Engineering Peruvian University of Applied Sciences Lima 33 Peru

Correspondence should be addressed to Christian Di Laura christiandilauragmailcom

Received 26 September 2015 Accepted 6 April 2016

Academic Editor Massimiliano Laddomada

Copyright copy 2016 Christian Di Laura et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

Today eavesdropping is becoming a common issue in the rapidly growing digital network and has foreseen the need for secretcommunication channels embedded in digital media In this paper a novel steganography technique designed for StandardDefinition Digital Television (SDTV) H264AVC encoded video sequences is presented The algorithm introduced here makesuse of the compression properties of the Context Adaptive Variable Length Coding (CAVLC) entropy encoder to achieve a lowcomplexity and real-time inserting method The chosen scheme hides the private message directly in the H264AVC bit streamby modifying the AC frequency quantized residual luminance coefficients of intrapredicted I-frames In order to avoid errorpropagation in adjacent blocks an interlaced embedding strategy is applied Likewise the steganography technique proposed allowsself-detection of the hidden message at the target destination The code source was implemented by mixing MATLAB 2010 b andJava development environments Finally experimental results have been assessed through objective and subjective qualitymeasuresand reveal that less visible artifacts are produced with the technique proposed by reaching PSNR values above 400 dB and anembedding bit rate average per secret communication channel of 425 bitssec This exemplifies that steganography is affordable indigital television

1 Introduction

Over the past decade digital media have become without adoubt part of our daily lifeThe recent technological achieve-ments in electrical and communications engineering havemade feasible the existence of a digital connected world andenormous amounts of data exchange However this situationhas lately exposed digital media to eavesdropping coun-terfeit and even sabotage turning it into a major securityproblem In pursuit of secure communications data cipher-ing techniques have been typically applied and preferredrather than other hiding methods Notwithstanding thesestrategies have failed to protect the reliability of the messageitself and have made it completely vulnerable to maliciousattacks In view of this data concealment techniques suchas steganography have increased their relevance and catchedthe attention of researchers ldquoSteganographyrdquo from the Greeksteganos graphos means ldquocovered writingrdquo and is consideredthe art of hiding secret data into a carrier medium so as

to convey the confidential message in such a way that itcannot be noticed or detected [1] A basic steganographyframework consists of an embedder and a detector The firstintroduces the secret data using a special algorithm into thecover work generating the so-called stegoobject and thelatter is responsible for extracting the hidden message usingthe right algorithm

As of this writing there have been plenty of researchworks on the use of media files such as audio images andvideo as cover files for steganography However they arenot designed and applied to current leading technologies forexample digital television

Nowadays digital terrestrial television broadcasting hasbeen successfully implemented This technology implies thecoding and compression of a high quality audio and videosource using a digital video standard and subsequent digitaltransmission In order to reach digital television severalvideo standards have been proposed H261 [2] H262 [3]H264AVC [4] and the latest H265High Efficiency Video

Hindawi Publishing CorporationInternational Journal of Digital Multimedia BroadcastingVolume 2016 Article ID 6950592 9 pageshttpdxdoiorg10115520166950592

2 International Journal of Digital Multimedia Broadcasting

Coding (HEVC) [5] (which is still under development fordigital television) H264AVC the standard most used inpractice has introduced several improvements on the hybridvideo encoding paradigm by adding new coding techniquesin the spatial transform and residual domain in search ofcompression These changes are mainly seen in the use ofnew prediction schemes for intra- and interprediction ofvariable block size within macroblocks of a new transformcore of the first entropy coding tools that take into accountthe importance of the context of the data being codedand of adaptive strategies to reduce the bit rate The firsttelevision specification using H264AVC was the BrazilianDigital Television Standard (SBTVD-T) and it is expectedthat all digital terrestrial television standards will start usingit in the coming years

Under these circumstances the lack of steganographytechniques and applications designed for digital televisionbecomes a field of much interest Currently there are nosteganography techniques designed for digital televisionFurthermore similar implementations are limited to theuse of audio [6] and images [7 8] as cover media Themost important related works consist of academic articles onwatermaking (similar to steganography with the importantdifference that the data to be concealed is related to thecover file) Nevertheless only a few of them are suitable fordigital television broadcasting where a real-time mediumembedding-strength no bit rate increasing technique isdesired

In [9] for example an interesting watermarking tech-nique was applied by making use of the coding propertiesof the CAVLC and inter-prediction The secret message isembedded in the sign bit of the high frequency coefficientscoded as trailing ones and in fixed bit codes used intransform blocks of inter-predicted frames Although theoriginal bit rate is not changed this technique does notconsider the error propagation originated by intra-predictionand leaves the possibility of embedding more data into thevideo stream

On the other hand in [10] a self-detection randomwatermarking technique using a key-dependent strategy wasproposed The odd and even characteristics of the quantizedresidual coefficients are used to hide and identify watermark-ing bits As a result a bit rate increment of less than 1 isachieved using this algorithm This proposal does not takein account the changes on the local context properties andthe final perception of the viewer as it selects the watermarkcoefficient randomly so creating possible visual artifacts

In [11] the authors suggested the use of a perceptualanalysis in order to create a robust watermarking techniquefor H264AVC video This method uses a human visionmodel created by Watson [12] that embeds the watermarkbits in the quantized residual luminance coefficients of whichthe quantization step size is at least changed by one Bythis procedure embedding capacity is gained Even if thehuman perception is considered this technique increases thebit rate by more than 5 and demands more computationalresources In addition the perception model from Watsonwas initially thought for Joint Pictures Expert Group (JPEG)still images based on the Discrete Cosine Transform (DCT)

H264AVC uses another type of transform the IntegerCosine Transform (ICT) of which the properties tend to bethose of the discrete cosine transform but are indeed notcompletely the same

Finally the works cited use only objective quality mea-sures in order to validate their simulation results and theembedding strength of their techniques However they donot consider the human real perception of the resultantencoded video sequences Hence subjective qualitymeasuresmust be considered in future works

This paper aims to present a real-time low-complexityself-detection and reduced-error-propagation-orientedalgorithm which maintains the bit rate of the stegovideosequences for digital television For this purpose secretmessages are hidden in the high frequency coefficientsof intrapredicted luminance blocks in an interlaced wayand making use of the properties of the entropy coder Inaddition simulation results are analyzed with an objectiveand subjective perception criterion

The paper is organized as follows Section 2 brieflypresents the extraction and insertion process of video fromdigital television and discusses details of the embeddingdetection and enhanced features of the steganography pro-posal Section 3 shows the experimental results fromobjectiveand subjective quality measures Finally Section 4 describesthe most remarkable conclusions and provides guidelines forfuture works

2 Scheme Proposed

In order to apply steganography to H264AVC videosequences from digital television these need to be separatedfrom the raw digital television bit streamTheway the schemeproposed solves this issue is part of the preprocessing stageof the final algorithm The digital television stream containsvideo audio control and synchronization information ofdifferent television programs specially packed using theMPEG-2 Systems specification also known as transportstream (TS) [13 14] The TS standard explains how differenttelevision programs composed of audio and video packetsboth considered Packetized Elementary Stream (PES) con-trol information and Programs Specific Information (PSI)composed of the ProgramAssociationTables (PAT) ProgramMap Tables (PMT) and timing information provided by theProgram Clock Reference (PCR) are alternately addressedand labeled with a Packet Identifier (PID) and joined into onecompliant bit stream The selected scheme extracts the Ele-mentary Stream (ES) corresponding to the H264AVC videoof one of the television programs The insertion mechanismoperates in the same way but in reverse order A detaileddescription of both processes including the embedding ofthe secret message using steganography and new featuresof the proposal is depicted in Figure 1 It is important toemphasize again that there are currently no state-of-the-artsteganography techniques designed for digital television

21 H264AVC Detection from MPEG-2 TS A MPEG-2 TSis composed of 188-byte-long transport packets that containvideo audio data and control information of the television

International Journal of Digital Multimedia Broadcasting 3

PAT (0x00)

PMT PID (0x120)

PMT PID (0x103)

PMT PID (0x1FC8)

PMT PID (0x0101)

PCR (0x0120)

ES audio (0x132)

ES video H264 (0x0121)

Steganograpyprocess

Stego ES video (0x0121)

PMT PID (0x0101)

Encryption (optional)ProtocolBinary

message

Repacking

streamStegotransport

Sync byte

(H222) signalstream

Transport

Figure 1 Overview of the scheme proposed

programs All of these signals are multiplexed in one bitstream The first step in the entire preembedding process isto extract the video ES of the chosen program from the TSThe procedure works as follows

(i) Synchronizing with the transport stream The decoderfetches the synchronization byte in the TS of whichthe value is always 0x47 base 16 (HEX) Each synchro-nization byte must be 188 bytes spaced without con-sidering any additional data introduced for ForwardError Correction (FEC) before its pattern is repeated

(ii) Program Specific Information (PSI) There are tworelevant tables that describe the instantaneous struc-ture of the transport stream and are sent as part ofthe control information The first table is the PATwhich is identified when the PID value equals 0x000HEX as specified in the standard This table showsthe total programs being carried in the TS and theirrelated PID labels The second table is the PMT Thistable allows associating each PID of an ES suchas video audio or additional control data with itscorresponding television program

(iii) Accessing a program After the PIDs of all ES havebecome known it is possible to have access to anyprogram carried by the TS and therefore to distin-guish between different ES Assuming that the videoand audio PID values are 0x121 HEX and 0x132 HEXrespectively the stream packets of which the PIDvalue equals 0x121 HEX will be assembled into onevideo ES and supplied to the next decoding phase (thesamewill happen if the TS is fetched for the audio ES)

After splitting the TS one video ES is chosen forembedding purposes Normally the first video frame to bedecoded is called Intraframe (or I-frame) as it is predictedfrom samples previously decoded belonging to the sameframe In addition the TS bit stream format allows thedetection of I-frames by using a special byte known as randomaccess indicator In some cases the first frame will have two

partitions or slices the top is predicted by intraprediction andthe bottom by interprediction (also known as P-slice) as ispredicted from samples previously coded belonging to otherframes or slices

22 Data Hiding Procedure in the H264AVC Sequence TheI-frame where the secret data will be embedded is dividedinto 16 times 16-pixel regions defined as macroblocks (this wayof splitting is done for coding and compression efficiencypurposes)Depending on the intraprediction scheme selectedduring the coding process they are once more divided into4 times 4 or 8 times 8 block sections H264AVC offers three typesof intraprediction modes mainly intra-16 times 16 intra-4 times 4and intra-8 times 8 (used in high profiles and some televisionapplications) Generally intra-16 times 16 is chosen for frames orslices (part of a frame) with predominant smooth areas andthe intra-4times 4 for the ones that contain rich amount of detailsIntra-8 times 8 is an especial case and will not be considered inthis work Likewise since human eyes are more sensitive tochanges in flat areas only intra-4 times 4 macroblocks have beenchosen for data embedding

Take into account an I-frame containing a macroblockregion that is coded using intra-4 times 4 prediction mode anddivided into block sections of 4 times 4 dimensions (furtherknown separately as ldquoTrdquo blocks) Now define ldquo119875(119894 119895)rdquo asthe pixel values of ldquoTrdquo and ldquo119868(119894 119895)rdquo as the correspondingintraprediction block where ldquo119894rdquo and ldquo119895rdquo stand for the discreteindexes of the rows and columns of the frame respectivelyTheir difference or residual matrix is labeled as ldquo119877(119894 119895)rdquo

H264AVC uses the ICT based on a 4 times 4 dimen-sional and standardized core for coding luminance andchrominance residual blocks Thus ldquo119877(119894 119895)rdquo is transformedscaled and quantized warranting the orthogonality andorthonormality properties during the process as explainedin [15] generating ldquo1198621015840(119906 V)

119905rdquo Consider ldquo1198621015840(119906 V)

119905rdquo as the

residual quantized ICT coefficients of block ldquoTrdquo ldquo119878(119906 V)119905rdquo

as the steganography data (in binary form) to be embed-ded ldquo11986210158401015840(119906 V)

119905rdquo as the residual quantized ICT coefficients

to which steganography has been applied ldquoTrcrdquo as an

4 International Journal of Digital Multimedia Broadcasting

Intraframe

Split in 16 times 16MBs

Current MB

Intraprediction

Transform

Quantization

CAVLC H264 stream

Binarymessage

BlockT R(i j)

I(i j)

P(i j)+

+

++

+

minus

S(u )t

C998400998400(u )t

C998400(u )t

0 1

2 3

4 5

6 7

8 9

10 11

12 13

14 15

Intra 4 times 4 MB t = 0 4 8 12

Figure 2 Steganography in H264AVC

experimental-achieved threshold and ldquo119906rdquo ldquoVrdquo and ldquo119905rdquo asdiscrete indexes within block ldquoTrdquo The embedding strategyworks as follows

If ldquo119878(119906 V)119905= 0 and 119862

1015840(119906 V)119905gt Trcrdquo then

11986210158401015840

(119906 V)119905=

1198621015840(119906 V)119905minus 1 if 1198621015840 (119906 V)

119905mod 2 = 1

1198621015840(119906 V)119905 if 1198621015840 (119906 V)

119905mod 2 = 0

119906 = 0 1 2 4 V = 0 1 2 4 119905 = 0 4 8 12

(1)

where ldquo1198621015840(119906 V)119905mod 2rdquo returns the remainder after

ldquo1198621015840(119906 V)119905rdquo is divided by ldquo2rdquoThe sign of the result is the same

as the divisorIf ldquo119878(119906 V)

119905= 1 and 119862

1015840(119906 V)119905gt Trcrdquo then

11986210158401015840

(119906 V)119905=

1198621015840(119906 V)119905 if 1198621015840 (119906 V)

119905mod 2 = 1

1198621015840(119906 V)119905minus 1 if 1198621015840 (119906 V)

119905mod 2 = 0

(2)

According to Figure 2 the steganographic algorithm pro-posed works at a quantization level because the sending is alossless operation and information embeddedwill not changein reception The embedding algorithm consists in separat-ing the embeddable block from closer neighbors inside amacroblock and leaving one macroblock of space betweenembeddable macroblocks The separate spaces where themessage is to be hidden are used to reduce visible artifactsso that subjective quality is not degraded

There is an exceptional case where the self-collusionattack must be avoided This is accomplished whenldquo1198621015840(119906 V)

119905= Trc + 1 and 119878(119906 V)

119905= 0rdquo The expected result

using (1) would be the same coefficient reduced by one levelHowever at the receiver side the original bit of the hiddenmessage will be lost and the recovering processes will fail Forthis reason and only for this special case the coefficient levelis incremented by one It should be noted that the binarymessage being embedded could be distributed to make

this singular condition less probable In addition the stegoTS produced does not differ from the original bit streamstructure This is achieved since the bit stream alienationand Program Clock Reference (PCR) are respected andpreserved The algorithm makes use of CAVLC features forlimited control of bit rate The target is to maintain the sameH264 sequence length or decrease some bytes in order to beable to insert the stego-H264 sequence in the same amountof TS packets

Four facts must be considered so as not to increase the bitrate

(i) Not hiding in zero macroblocks(ii) Not choosing Trailing Ones coefficients (T1s) as they

have a defined code(iii) Not choosing zero coefficients in any nonzero mac-

roblock as the length will be increased(iv) Reducing luminance ICT coefficients in one level that

will produce a smaller length than the original

Finally the resultant I-frame is repacked into the TS andso ready to be transmitted

23 Hidden Message Extraction The retrieval of the hiddenmessage is fast and simple By applying the same processshown in Figure 2 to extract an I-frame and after entropydecoding the H264AVC video embeddable macroblocksare chosen and ready to extract the hidden message The waythe secret message is recovered bit by bit is given in

119878 (119906 V)119905=

0 if 11986210158401015840 (119906 V)119905mod 2 = 0

1 if 11986210158401015840 (119906 V)119905mod 2 = 1

(3)

24 Embedding Protocol and Ciphering With the purposeof enhancing the security of the embedded message and

International Journal of Digital Multimedia Broadcasting 5

warranting self-detection at the target destination the pro-posal ciphers the hidden message prior to the steganographytechnique using the RC4 algorithm [16] and wraps it into aspecial designed protocol which ensures self-detection at theend-peer RC4 which was chosen due to its programmingsimplicity and performance [17] encrypts the message byusing a private key Thus if the steganographic algorithm isbroken the enemy would have to know the secret key andcipher algorithm

To sum up in this section a novel and low-complexitysteganographic algorithm for digital televisionwas presentedThe proposed scheme first decodes the TS from digital tele-vision extracting H264AVC video sequences Afterwardsthese are decoded and searched for intrapredicted lumaregions where the secret and previously ciphered messagesare hidden into the ICT high frequency residual coefficientsusing an interlaced embedding strategy and exploiting theproperties of the CAVLC entropy encoder depicted in thesteganographic method Finally the resultant stegovideosequences are repacked into the TS stream

3 Experimental Results

A software simulation of the proposed steganographytechnique was implemented by mixing MATLAB 2010 band Java development environments The most importantachievements of the program are the MPEG-2 Systemsand H264AVC integration a proprietary code that meetsstandards In addition a novel CAVLC entropy encoderscheme using a lookup-table strategy extracted from [18] isaccomplished This allowed improving the overall encodingand decoding speed for residual luminance coefficients Allexperiments are conducted using a transport stream samplefrom a typical free-to-air television channel from whichthree different cover video sequences from the daily programlist were extracted The selected video sequences differed inthe changes within the scenes and were addressed as staticmoderate and highly variant

Static video is characterized by slow and almost nullmotion in the scene where the focus of the camera iscommonly centered on capturing one objectThemoderatelyvariant video is characterized by some scene changes andhighly variant video is distinguished by the fast alternationof uncorrelated scenes In order to prove the embeddingcapacity of each video sequence and the amount of distortionintroduced three different randomly selected and different-length secretmessageswere prepared Simulation resultswerecompared using two different quality measures One of themwas the objective quality measure known as Peak Signal toNoise Ratio (PSNR) [19] which is a widely adopted methodused in engineering to measure the amount of distortionintroduced by compression comparing the original with thestegovideo frames in a pure mathematical way and the otherwas the subjective perceptual criterion known as SubjectiveDifference Grade (SDG) based on ITU-R BS1116 [20] Thelatter was successfully implemented in [21] and is now appliedto digital television

Unlike PSNR SDG considers the viewerrsquos opinion bytaking a survey The SDG procedure works as follows There

5

4

3

2

1

annoyingVery

Annoying

annoyingSlightly

not annoyingPerceptible but

Imperceptible

Figure 3 Subjective test scores

are three different digital television sequences labeled as ldquoArdquoldquoBrdquo and ldquoCrdquo ldquoArdquo is always the original sequence while ldquoBrdquoand ldquoCrdquo are randomly scrambled and one of them containsthe original and the other the stegosequence The viewer isasked to watch ldquoArdquo and informed that this is the originalone and then ldquoBrdquo and ldquoCrdquo without telling them about thecontent of both sequences Then the viewer is requested topunctuate the distortion perceived in ldquoBrdquo in relation to ldquoArdquoand ldquoCrdquo in relation to ldquoArdquo The possible marks that can bechosen for this subjective test are detailed in Figure 3 where5 is assigned when no difference is perceived and 1 when it isvery annoying

After the survey the SDG value is calculated as a specialquality measure function In this paper the SDG valueis obtained from the difference between the punctuationassigned to the stegofile and the original as shown in

SDG = ScoreStego minus ScoreOriginal (4)

If the SDG value becomes highly negative it means thatnoticeable distortion has been introduced by the steganogra-phy technique If it turns positive or close to zero howeverit can be inferred that perceptual degradation has not beenperceived and subjective quality is better

31 Simulation Conditions The sequences used during thesimulations had the following common properties All arecoded in H264AVC high profile use the CAVLC entropycoder and a frame rate of 30 fps and have a spatial resolutionof 720 times 480 pixels In addition the quantization parameterof each macroblock is not fixed and varies adaptively withinthe different frames as well as the scan order which couldbe interlaced or progressive Furthermore each test sequencelasts 15 seconds The opinion survey consists of 27 differentquestions with a sample of 30 people aged between 20 and30 years in a well-illuminated environment These tests con-solidate three different message lengths and three differentsteganography techniques The steganographic methods arecomposed of [10] an earlier version of the proposed tech-nique without considering the interlaced strategy explainedin Figure 2 and with the main difference that the discreteindex ldquo119905rdquo had the range 119905 = 0 1 15 and the final proposalLikewise the method complexity of [10] and the proposalis depicted in Table 1 The message lengths are estimated asa percentage of the maximum embedding capacity of eachvideo excerpt using the technique in [10]

32 Simulation Results Figure 4 depicts the PSNR of frame30 for the method proposed in [10] for the steganographicmethod without interlaced strategy and for the final pro-posal tested with three different messages lengths and three

6 International Journal of Digital Multimedia Broadcasting

minus21

minus18

minus15

250 400 550 700

SDG

scor

e

(Byte)

NoorkamiProposedFirst test

(a) SDG scores of static video

4005

4045

4085

4125

4165

4205

250 400 550 700

PSN

R (d

B)

(Byte)

NoorkamiProposedFirst test

(b) PSNR values of static video

minus03

minus02

minus01

0

01

02

03100 200 300

SDG

scor

e

(Byte)

NoorkamiProposedFirst test

(c) SDG scores of moderately variant video

402

407

412

417

100 200 300

PSN

R (d

B)

(Byte)

NoorkamiProposedFirst test

(d) PSNR values of moderately variant video

minus06

minus05

minus04

minus03

minus02

minus01

050 150 250

SDG

scor

e

(Byte)

NoorkamiProposedFirst test

(e) SDG scores of highly variant video

401

406

411

416

421

50 150 250

PSN

R (d

B)

(Byte)

NoorkamiProposedFirst test

(f) PSNR values of highly variant video

Figure 4 SDG scores and PSNR values achieved with the first test proposed algorithm and [10]

International Journal of Digital Multimedia Broadcasting 7

Table 1 Comparative table of the technique proposed and [10]

Technique Hidingcomponent Applied to Security Computational

complexity

MaximumembeddingbitsMacroblock

Perceptualawareness Key feature

Reference [10] CAVLC codeword

IntrapredictedI-frames(luminancesamples)

Key-dependentstrategy basedon a public keyextracted fromthe localmacroblockfeatures and aprivate keyowned at thetargetdestination

Low One

No thealgorithm canchange low andhigh frequencyresidual ICTquantizedcoefficients

Hidden datarandomness dueto a key and localmacroblockderivativesdependentalgorithm

Proposed CAVLC codeword

IntrapredictedI-frames(luminancesamples)

The secretmessage isciphered using aprivate key [16]prior to theembeddingprocess anddivided intobyte units

Lower (securitydoes not rely onthe localmacroblockfeatures whichdemands highercomputationalresources)

Four

Yes thealgorithm isdesigned tochange onlyhigh frequencyresidual ICTquantizedcoefficients andin an interlacedway in order totry to avoidperceptualdegradation anderrorpropagation

Deepunderstanding ofthe codingproperties of theCAVLC entropyencoder for bitrate control andrapid embeddingalgorithmpurposes

different digital television sequences respectively The sim-ulation results show that the PSNR values tend to decreaseas the embedded message size increases However all thePSNR values are above 40 dB which is an acceptable rangefor steganography techniques in the H264AVC In additionthis denotes that the objective quality of the stegosequencesis good and that few errors are introduced by the embeddingtechniquesThe results also illustrate that better PSNR valuesare achieved by the technique proposed for moderatelyvariant and static video slightly outperforming both [10] andthe technique without interlaced strategy Figure 5 depictsframe 30 of the original video frame the secret data and thestegoframe for the different types of video

Dissimilar results from the PSNR analysis are obtainedwith respect to the SDG values These are shown in Figures4(a) 4(c) and 4(e) and are tested under the same simulationconditions depicted in Figures 4(b) 4(d) and 4(f) First ofall the static sequence clearly shows that as the embeddingdata length increases the SDG value rapidly decreases dueto high light intensity and fewer camera movements becausethe camera is focusing on the center of the scene where thepresenter is talking However it is a fact that the human eyegets easily used to quasi-invariant scenes and smaller changesare quickly recognized On the other hand it is interestingto note that some highly and moderately variant excerptsgot positive or near to zero SDG values during the surveyThis can be only justified by the movement between thescenes which becomes an important topic for the analysis

since changes mask the slight distortions introduced bythe steganography technique and increase the complexityof the viewerrsquos choice For this reason the PSNR is notsufficient criterion to decide if the designed steganographytechnique outperforms the other ones For example the staticvideo has acceptable PSNR values but it is the viewerrsquosopinion that it is of bad quality In this context it should bepointed out that the steganography technique of the proposalslightly outperforms the rest of hiding methods in the SDGsimulations The technique proposed does not change theICT coefficient randomly as in [10] Furthermore the freespaces left between embeddable macroblocks and the specialstrategy to select the embeddable blocks manage to reducevisible artifacts and improve the objective and subjectivequality of the stegosequences on average Finally Table 2shows the embedded bit rates reached by the proposal and by[10] for several video sequences fromdigital television as wellas the bit rate increase in percentage related to the originalvideo excerpts

It should be noted that the technique proposed reachedan average embedded bit rate of approximately 425 bitssecper secret communication channel without increasing theoriginal video bit rate but slightly decreasing it instead andso exemplifying the bit rate control of the proposal overthe CAVLC encoder Table 2 is relevant as it describes thepossible bandwidth limits that applications of steganographyin SDTV will face

8 International Journal of Digital Multimedia Broadcasting

(a) Original static video frame (b) Stegoframe of static video

(c) Original moderately variant video frame (d) Stegoframe of moderately variant video

(e) Original highly variant video frame (f) Stegoframe of highly variant video

Figure 5 Stegovideo frames excerpts after hiding a secret message (secret data of 80-byte length ldquothe two companies are at the forefront ofa tantalizing wireless communications concept that has proved hard to produce on a big scale reduce cellphone costs by relying on strategicallyplaced Wi-Fi routers And when there are no routers available fall back on the traditional cellular networkrdquo)

Table 2 Embedded bit rate in bitssec for the technique proposedand that of [10]

Testsequence

Reference [10] ProposedEmbeddedbit rate inbitssec

Bit rateincrease in

Embeddedbit rate inbitssec

Bit rateincrease in

Highlyvariant 402 033 400 minus016

Moderatelyvariant 544 051 464 minus002

Static 384 079 416 gtminus001

4 Conclusion

In this paper a novel real-time affordable and compresseddomain steganography technique for SDBTV digital televi-sion sequences is discussed The secret message is hidden in

the high frequency luminance ICT coefficients of intrapre-dicted 4 times 4 macroblocks using an interlaced embeddingstrategy Furthermore self-detection at the target destinationand enhanced security features are achieved by applying aspecial embedding protocol and ciphering the secret dataprior to the steganographic process

Simulation results show that the technique proposedmaintains a good subjective quality of the stegosequence andthat the PSNR analysis is not sufficient criterion to assurethat the perceptual quality will not be degraded during thesteganographic process It is very important to take intoaccount the viewerrsquos real perception This was proved bythe different results achieved between the SDG and PSNRmeasure Finally it is worth mentioning that a special careshould be provided to static sequences where slight changesmay cause annoying visual artifacts

A first steganography technique designed for digitaltelevision is here presented In future works the schemeproposed will be based on intrapredicted 8 times 8 macroblocks

International Journal of Digital Multimedia Broadcasting 9

with an adaptive brightness algorithm to reduce the embed-ding capacity in static sequences and thus preserve subjectivequality In addition new standardized algorithms for anobjective quality measure are under investigation

Competing Interests

The authors declare that they have no competing interests

Acknowledgments

The authors would like to thank the Peruvian Instituteof Radio and Television (IRTP) for providing the digitaltelevision sequences that the work was based on

References

[1] I Cox M Miller J Bloom J Fridrich and T Kalker DigitalWatermaking and Steganography Morgan Kauffmann 2008

[2] ITU-T ldquoVideo codec for audiovisual services at p times 64 kbitsrdquoRecommendation H261 ITU-T 1993

[3] ISOIEC 13818-2 and ITU-T Rec H2620 ldquoInformationtechnologymdashgeneric coding of moving pictures and associatedaudio information videordquo ISOIEC JTC 1 and ITU-T 1995

[4] ISOIEC and ITU- T ldquoAdvanced video coding for genericaudio-visual servicesrdquo ISOIEC 14496-10 and ITU- T Recom-mendation H264 ISOIEC JTC 1 ITU-T 2010

[5] G J Sullivan J-R OhmW-J Han and TWiegand ldquoOverviewof the high efficiency video coding (HEVC) standardrdquo IEEETransactions on Circuits and Systems for Video Technology vol22 no 12 pp 1649ndash1668 2012

[6] G Nehru and P Dhar ldquoA detailed look of audio steganographytechniques using LSB and genetic algorithm approachrdquo Inter-national Journal of Computer Science Issues vol 9 pp 402ndash4062012

[7] K-C Chang C-P Chang P S Huang and T-M Tu ldquoAnovel image steganographic method using tri-way pixel-valuedifferencingrdquo Journal of Multimedia vol 3 no 2 pp 37ndash442008

[8] Y-F Sun D-M Niu G-M Tang and Z-Z Gao ldquoOptimizedLSB matching steganography based on Fisher informationrdquoJournal of Multimedia vol 7 no 4 pp 295ndash302 2012

[9] B Mobasseri and Y N Raikar ldquoAuthentication of H264streams by direct watermarking of CAVLC blocksrdquo in SecuritySteganography and Watermarking of Multimedia Contents IXvol 6505 of Proceedings of SPIE The International Society forOptical Engineering San Jose Calif USA January 2007

[10] M Noorkami and R M Mersereau ldquoCompressed-domainvideo watermarking for H 264rdquo in Proceedings of the IEEEInternational Conference on Image Processing (ICIP rsquo05) vol 2pp 890ndash893 Genova Italy September 2005

[11] M Noorkami and R M Mersereau ldquoA framework for robustwatermarking of H264-encoded video with controllable detec-tion performancerdquo IEEE Transactions on Information Forensicsand Security vol 2 no 1 pp 14ndash23 2007

[12] A B Watson ldquoDCT quantization matrices visually optimizedfor individual imagesrdquo inHuman Vision Visual Processing andDigital Display IV vol 1913 of Proceedings of SPIE pp 202ndash216San Jose Calif USA September 1993

[13] ISOIEC and ITU-T ldquoInformation technologymdashgeneric codingof moving pictures and associated audio information sys-temsrdquo ISOIEC 13818-1 and ITU-T Recommendation H2220ISOIEC JCT 1 ITU-T 2006

[14] J Arnold M Frater and M Pickeking Digital TelevisionTechnology and Standards John Wiley amp Sons 2007

[15] I E RichardsonH264 and MPEG-4 Video Compression VideoCoding for Next-Generation Multimedia John Wiley amp SonsNew York NY USA 2003

[16] W Stallings The RC04 Stream Encryption Algorithm 2005httpcsespsueduafaruqueit6833RC4pdf

[17] S O Sharif and S P Mansoor ldquoPerformance analysis ofstream and block cipher algorithmsrdquo in Proceedings of the 3rdInternational Conference on Advanced Computer Theory andEngineering (ICACTE rsquo10) pp V1522ndashV1525 Chengdu ChinaAugust 2010

[18] Y Yi and B C Song ldquoHigh-speed CAVLC encoder for 1080p60-Hz H264 codecrdquo IEEE Signal Processing Letters vol 15 pp891ndash894 2008

[19] S Winkler and P Mohandas ldquoThe evolution of video qualitymeasurement from PSNR to hybrid metricsrdquo IEEE Transac-tions on Broadcasting vol 54 no 3 pp 660ndash668 2008

[20] ITU-R ldquoMethods for the subjective assessment of small impair-ments in audio systems includingmultichannel sound systemsrdquoITU-R Rec BS1116-1 1997

[21] G Kemper and Y Iano ldquoAn audio compression method basedon wavelets subband codingrdquo IEEE Latin America Transactionsvol 9 no 5 pp 610ndash621 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of