IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 Blind Co lor Image Steganography in Spatial Domain F. A. Abdullatif, W. A.Shuk ur Departme nt of Computer Science , College of Education I bn Al-Hatham, Unive rsity of Baghdad. Recei ved in December 19 2010 Accepted in Feb. 8 2011 Abstract With wide sp read of internet, and increase the p rice of information, st eganogr aphy become very imp ortant to communication. Over many y ears used different types of digital cover to hide information as a cover channel, ima ge from imp ortant digital cover used in st eganography because widely use in internet without suspicious. Since image is frequently comp ressed for storing and transmission, so steganogr aphy must counter the variations caused by loss compression algorithm. This p ap er describes a robust blind image st eganogr aphy , the prop osed method embeds the secret message without altering the quality by sp ray ing theme on the blocks in the h igh ord er bits in color chann el such blue "without altering the human vision sy stem" and make them integral to cover. This method dep ends on constant sequence sp read sp ectrum method and survives loss comp ression image like JPG. Keyword: information hidin g, color image steganohrapy , sp read sp ectrum. Introduction The prevalence of multimedia data in our electronic world exp oses a new avenu e for communication usin g digital steganography . Stegano graphic techniques are useful to convey hidden information by using various types of typically-transmitted multimedia data as cover for concealed communication. The inability to detect the hidden data, p erceptually or by comp uter analy sis, is p aramount for surreptitious op eration[1][2]. There are many applications for techniques that embed information within digital images. Stegano grp hy where the original cover si gnal is need ed to reveal the hidden information are known as cover escrow. In many applications it is not p ractical to require the possession of the unaltered cover si gnal in order to extract the hidden information. M ore p ragmatic methods, known as blind or oblivious schemes, allow direct extraction of the embedded data from the modified si gn al without knowledge of the original cover. A block diagr am of a blind ima ge steganographic sy st em is depicted in fig 1. A message is embedded in a digital image by the stegosy stem encoder. The resulting st egoima ge is transmitted over a channel to t he receiver where it is p rocessed by the stegosystem decoder. IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 During transmission the st egoima ge can be monitored by unintended viewer who will notice only the transmittal of the innocuous image without discovering the existence of the hidden message. It's known that any natural image has b ig amount of sp atial redundancies for this reason continues tone image comp ression exist to remove those redundant information and represent the data in more efficient form suitable for st oring and transmitt ing [3]. Although basic image comp ression st ander like JPG is based on DCT , but recently new image comp ression st ander has been p rop osed that is based on wavelet transform, but lik e any transform domain approach, DCT[4],wavelet[5]domain data hiding, demands high comp utational comp lexity that is disadvantage for real time data embedding and retrieval p rocess. M oreover, control on image v isual quality and simple low cost hardware realization of transform domain data hidin g is not alway s easy Sp read Sp ectrum Image Stegano graphy , is a data-hidin g/hidden-communication method that uses digital imagery as a cover signal. Sp read Sp ectrum Image Stegano graphy p rovides the ability to hide and recov er, error fr ee, a si gnificant quantity of information bits within digital images, avoiding detection by an observer. Furt hermore, Sp read Sp ectrum Image Steganography is a blind sch eme because the original i mage is not needed to extract the hidd en information. The prop osed recipient need only p ossess a key in order to reveal the hidden messa ge[6]. Proposed method In this section, we will describe the blind st eganography method that survives with loss comp ression, the main idea behind method that the joint p hotography exp ert group "jp eg" comp ression algor ithm loss some image information, but not change the ima ge quality . The jp eg algorithm is a loose algorithm this will p roduce small amount of visual d istortion, and will introduce error in the image data but this error data still nearest from original image data to avoid visual chan ge on image because lar ge change in data ( gray scale to any color band) make v isual distortion. For that we used 24 bitmap image as a cover and emb edded secret data in sp atial domain in 8*8 block in one chann el values and leaving the other chann el un chan ged then the st ego image is st ore in jpeg format In this method we use a table of 256 entries and divide the table to several p eriod, all p eriods are the same len gth as in table 1 and each center of period take 0 and 1 resp ectively. In the embeddin g p rocess, each block 8*8 from cover ima ge used to emb edded on e secret bit by adjusting the value of cover ima ge data to center of p eriod closed to secret bit, the p eriod and center of p eriod as in table 1 st ego table, example to exp lain embedded p rocess, let the secret bit is 0 and the valu e of blue chann el is 25 "we use b lue channel to embedded because human vision sy stem HVS is less sensitive to blue change than the red or green " we change the 25 to the nearest p eriods that have 0 "secret bit in examp le" in center these periods 32 -47 or 0-15 t he center of the first is 39 and the difference is 14 and the center of the second p eriod is 7 and the difference is 18, then change the valu e to 39 because had less difference "we use this two p eriods because the secret b it 0 and the value of the blu e channel is 25 and the nearest p eriods have 0 in center these p eriod". IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 Each p ixel "channel valu es "have two p eriods or more depend on secret bits excep t the boundaries value near 0 or n ear 255 have only one ad just value and not need to comp ute and comp are p rocess. This p rocess will repeat for all 8*8 blocks, in the worth case the ch ange not e xceeded the distortion caused by jep g algorithm. Aft er receiving st ego image we assembly the 8*8 blocks in blue channel and read the value to know the period value and then the secret bits. Extra method will use to overcome the high d istortion in some pixel value that can app ear in jp eg algor ithm "like noise and rap idly lar ge chan ge in value" that the extracting come fro m block 8*8 then can measure the entropy of 0 and 1"number of 0 and 1 in block " and if the entropy of 0 more than, the secret bit in block is 0 or if entropy of 1 lar ger, then the secret bit is 1, my exp erimental test show that no block have the same entropy to 0 and 1 The purpose of use 16 as interval length We use 16 as p eriod length this make the change in value less than the HVS sensitive and be normal with the dist ortion of jp eg algorithm and also large to overcome the error counter in jpeg comp ression, and also to avoid p roducing p attern can use in att ack where the output 16 numbers and after comp ress can be any number from 256 numbers. The propose d embedded algorithm 1- Sp lit the cover image to t hree array s RGB "we use a true color 24bit image" 2- Take the blue array and use the first 8*8 block to store o if secret message is text or 1 if secret message is a file. 2- if st ego message is text then take next 8*8 block to st ore 0 for Arabic and 1 for English then use encode decode table that use 6 bits t o encode the letter with 64 different sy mbol "if second 8*8 block h ave 0 then take Arabic letter with sp ecial char table or if 1 then take English letter with special ch ar table and the outp ut is binary array Or if st ego message is a f ile encode the file to binary array 3- Compute if the cover file can st ore all stego message or not Size st =(length*high) / 64 + i*64  Where i =2 if used text message and i =1 if used file If binary message array is more than Size st then suggest to replace the cover ima ge with other image hav e larger dimension and return t o 1 5-Repeat until all secret bits are embedd ed i-read secr et bit ii-for each 8*8 block :- read the data of block and change it to the nearest p eriod center depend on table 1 and p rocess comp ute and compare" comp ute difference and comp are to choose less difference" IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 6- Put end of message sign if the text message is used, use EOF sign if a file used Arabic-Engl ish encode and decode table This table contains two fields and 64 r ecords one field for char and other for 6 bits binary code representing char and the last record in Arabic and English t able is used ed to t he end messa ge sign In embedded process comp are the char with first field and put the corresp onding binary in second field to binary array In the extracting p rocess check the binary with the second field and take the corresp onding char in the first field We use this table for more se curity through encode each char in b inary code differ ent in asscii and for mor e sp ace through each letter takes 6 bits rather than 8. Extracting In extracting we read the first block and know the embedded message either text or file and if text read the second b lock to know decode in Arabic or English t able then read each block and after reading the value of each block items determine the p eriod then the secret bit from table1 after comp uting the entropy of 0 and 1 in block, until reachin g to the sign of end message in text or EOF sign in file, examp le if the first item in block have 178 then it is in 176-191 p eriod and the secret bit is 1 Ex perime ntal result We have tested the p rop osed embedding method with many images and show in low frequency area "smooth area" that have ap p roximate identical pixels values the number of distortion "error" in jpeg algorithm is less than the other area "high frequency area "for that we can use block smaller than 8*8 b lock in this typ e of area to embedd the se cret bits and recover with error free se cret message. For more robust we use multi periods t o make the distributed for all image value range and the p attern from the move image cover value to the center of p eriod can alter in the comp ression p rocess and the value of distortion from comp ression p rocess is different and within the range of normal compression as shown in fig 2 the two images used as cover and fig 3,4 the histogr ams to each bands RGB of the two images in fig 2 resp ectively before comp ress, fig5 the two images after hidin g information and compress and receive the other side, fig 6,7 the two images histograms of fig 5, we can notice that the distortion of the blue channel that is used to hide information approximately same in red R and green G where the R and G that not changed only the dist ortion of compression happ en on it. We can also use two or three bands for emb eddin g p rocess with low comp ression valu e "approximately less 60" Also we can use block 4*4 if the co mpression value less t han 55 Conclusion In this p aper, we have p rop osed new robust blind st eganography methods by this method st ore secret message in on e band "use blue b and" in 8*8 blocks and can survive with IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 jpeg compression and any DCT base compression without sensitive of HVS and st ill hav e the same st atist ical p rop erty . Re ferences 1. Bender, W.; Gruhl, D.; M orimoto, N. and Lu, A.(1996) , Techniques for data hiding .IBM system journal,35(3-4). 2. Cox, I. J.; Kilian, J. and Shamoon, T. September (1996), secure sp read sp ectrum watermarkin g for ima ges, audio and video. Proceeding of IEEE international conference on image processing, Lausann e,Switzerland,III:243-246 3. Fridrich, (1998) , App lication of data hiding in digital image ,URL:http ://www.ssie.bingh amton.edu/~jirif 4. Riley M ary , J. and Richardson Iain, E.G.( 2007), Digital image and Video Communications, Art echHouse, Bosten,London. 5. Tsai, C.S., C. C, T. S., Chen ,( 2006), Sh arin g multip le secrets in digital ima ges, the journal of sy st ems and software. 6. Avcibas, I. M emon, N. and Sankur ,B.(2008), steganaly sis using ima ge quality metrics, Wiley -interscience . Table: (1) The stegano table 0-15 16-31 32-47 48-63 64-79 80-95 96-111 112-127 128-143 144-159 160-175 176-191 192-207 208-223 224-239 240-255 7 23 39 55 71 87 103 119 135 151 167 183 199 215 231 247 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 Fig. (1): a bli nd image steganographi c syste m IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 a b a b Fig.( 2): a and b is a BMP image use d as cover image a b Fig.( 3): a – red band histogram, b- gree n Band histogram and c- blue band Histogram for fig 2 a c IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 a b c Fig.( 4): a – red band histogram, b- gree n Band histogram and c- blue band Histogram for fig 2 b a b Fig.( 5): a i s the same of fig 2 a in the receiver side and b is the same of fig 2 b in the receiver side that use to extract message IBN AL- HAITHAM J. FOR PURE & APPL. S CI. VOL.24 (1) 2011 a b Fig.( 6): a – red band histogram, b- gree n Band histogram and c- blue band Histogram for fig 5a c a b Fig. (7 )a – red band histogram, b- gree n and histogram and c- blue band Histogram for fig 5 b c 2011) 1( 24المجلد مجلة ابن الهیثم للعلوم الصرفة والتطبیقیة طریقة عمیاء إلخفاء بیانات داخل صورة ملونة في مجال الصورة وسام عبد شكر ، فراس عبد الحمید عبد اللطیف ، جامعة بغداد ابن الهیثم -كلیة التربیة ،قسم علوم الحاسبات 2010كانون االول 19استلم البحث في 2011شباط 8قبل البحث في الخالصة .جدا "امهم "امخفیة أمر الأصبح نقل البیانات ، مع االنتشار الواسع لالنترنت والزیادة الكبیرة في قیمة المعلومات الصورة دتع، ت وٕارسالها في قنوات مخفیةرقمیة إلخفاء المعلوماعلى مر السنین استعملت أنواع مختلفة من األغطیة ال .دون أي شكوكمن من أهم هذه األغطیة الرقمیة النتشارها الكبیر في االنترنت وٕامكانیة تداولها تضغط عاد ق أخفاء البیانات قادرة على االحتفاظ بالبیانات مع هذا ائللخزن والنقل لهذا یجب أن تكون طر ةبما إن الصورة .الضغط دون من ن الطریقة المقترحة تخفي النص السريا اذ، هذا البحث طریقة جیدة إلخفاء البیانات في صورة یوصف مثل قناة ،صورةفي قناة من قنوات ال" البلوكات" تغیر بنوعیة بیانات الصورة وذلك برش النص على مجموعة من الكتل جعل الرسالة السریة متكاملة مع خصائص الصورة ممایر في الصورة یدون تغیمن اللون األزرق و الرسالة المخفیة صحیحة حتى بعد الضغط مع ىهذه الطریقة على طریقة نشر السلسلة الطیفیة الثابتة وتبقتعتمد " ضغط الخسارة"الفقدان