Indonesi an  Journa of  El ect ri cal Engineer ing  an d  Comp ut er  Scie nce   Vo l.   13 ,  No.   1 Jan uar y   201 9 ,   pp.  147~ 154   IS S N: 25 02 - 4752, DO I: 10 .11 591/ijeecs .v1 3 .i 1 .pp 147 - 154          147       Journ al h om e page http: // ia es core.c om/j ourn als/i ndex. ph p/ij eecs   Developi ng audio  data hid ing sche me using  random  sa mp le  bits  with lo gical  operat or s       Moham med  Hatem  Ali Al - Hooti 1 , T ohar i Ahm ad 2 ,  Supeno  Djanali 3   1,2,3 Depa rtment   o Inform at i cs,   In st it ut Te kno logi   Sepuluh  Nopem ber Ea st Java ,   Indone sia   1 Facul t y   of  Com pute Sc ie nc e an Inform at ion   T ec hnolog y ,   San a ’a  Univer si t y ,   Sa na’ a, Ye m en       Art ic le  In f o     ABSTR A CT    Art ic le  history:   Re cei ved   J ul   6 ,  201 8   Re vised  N ov   2 201 8   Accepte Nov  1 9 , 201 8       Sharp  deve lopment  progre ss   of  informati on  tech nolog y   has  aff e ct ed  m an y   aspe ct includin dat sec ur ity .   Thi is  bec ause   cl assifi ed  data  are   oft e n   tra nsferre d   bet w ee s y s te m s.  In  thi ca se ,   da ta   h idi ng  ex ists  to  p rote c such   dat a .   Som m et h ods  which  hav be en  proposed ,   howeve r,  are  not   y et  opti m a l   conc ern ing  th a m ount  of  the  sec ret   and  th qua li t y   of   th resul te stego  data .   In  thi pape r ,   we   expl ore   an  aud i fil as  the   m ed ium  to  ca rr y   th sec ret   d ata  which  has  b ee n   ext ra ct ed   int o   bi nar y .   B efo re   the  proc es beg ins,  the   cove is   conve rt ed  to  bi nar y   and  each  sam ple ’s  bit are   divi ded  int t wo  groups,    one  is  used  as  t he  locati on   of  t he  embedde 4   bit where as  the  sec ond  par t   loc a te the   two  bit tha are   ran dom l y   select ed  as  the   ke y .   The   expe riment a l   result have   val i dat ed  that  the   capac i t y   is  high  a nd  the re  is  no  muc impact  on  the  qua li t y .   Moreove r,  compare to  the   cur re nt  LSB  m et hods   the  sec u r i t y   is e xc ee ding l y   e nhanc ed .   Ke yw or d s :   Data hi ding   Data p r otect ion   Inform at ion  secur it y   Secret  data   Stegan ogra phy   Copyright   ©   201 9   Instit ut o f Ad vanc ed   Engi n ee r ing  and  S cienc e   Al l   rights re serv ed .   Corres pond in Aut h or :   Moh am m ed  H at e m  A li  A l - H oo ti   Dep a rtm ent o f Info rm at ic s,   In sti tut Te knol og Sepulu h N op em ber ,   Su r abaya,  S ukolil o,   60111,  E ast  Jav a ,  In done sia .   Em a il m oh _h a t84@y ah oo.com       1.   INTROD U CTION     Pr ese ntly m ultim edia  network   a nd  com m un ic at ion   is  s ha rp ly   pro gressi ng.  Di gital   com m un ic at ion   has  bec om so   sign ific ant  f or   m any  or ga nizat ion s.  It  s upports  eve daily   li fe  act ivities  and   c omm un ic at ion [1,  2].  F or  i ns ta nce we  t ran s fer  m edical   i m ages,  dig i ta au dio s,   m us ic vid eo s,   dig it al   book s,  et c.    This  tra ns m is sion  occ urs  on  public  networks  s uc a the  i nter net  w hich   m eans,  a nyone  ca easi ly     acce ss  them As  res ult,  tho s file m ig ht  be  sim ply  forg e or   m anipu la te d.   To  m ini m iz this  threat ,   inf or m at i on   secur it has  com to  protect   data  against  so m issues,  s uch   a piracy,  data  f orger y,  et [ 3,   4].  I this  fiel d,  data  hid i ng  act as   rece nt  a nd  well - kn own  re search   to pic  w hich  is   sp e ci fied  int t he  f ol lowing   top ic su c a ste ganogra phy  that  is  con cer n in m uch   to  protect   the  com m un ic at ion pri vacy   [5 6].   Stegan ogra phy  aim to  hid e   secret  m ess age  us in ha rm le ss  cov er,   so   no  ot her   par ty   can   dete ct   this   secur e m essa ge.   I the  past data  hid in s ta rted  in  co nventional  te ch ni qu e su c as  inv isi ble  ink s sp rea d - sp ect r um   co m m un ic at ion s,  a nd   co ver c ha nnel [ 3,   7].  Cu rr e ntly this  ca be   car ried  ou by  hid i ng  the   secret  inside  a nothe cov e file   [ 8].  This  m edium   m igh be   an   i m age,  a ud i o,   vi deo ,   or   te xt  [ 9 - 13] Waterm ark i ng   [14 - 16]   con ce al secret  inform ati on   su c as  sign at ures  within  any  m edia  without  hav i ng   pe rc eptual   distor ti on  [ 17, 18].   Audio  data  hi di ng   sc hem es  are  cl assifi ed  int tw ty pes:  lo ssless  an loss y.  The  lossless   al gorithm   consi ders  retri evin both  th secret  m essage  an the  ori gin al   co ve file Loss m et hod  cares  only   on  recoveri ng   t he   secret  inf orm at ion   [19].  Ther a re  m a ny  pa ram et ers   wh ic are  use to  e valua te   the   perform ance  of  au dio   ste ganogr ph sc hem e s.  N ot  al of   th e m   sh ould  be  m et since  each  im ple m entati on   m ay   Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2502 - 4752   Ind on esi a J  E le c Eng &  Co m Sci,   Vo l.   13 , N o.   1 Ja nu a ry 20 19   :   147     154   148   hav diff e re nt  pur po se s.   W can  su m m ariz so m e   sign ific ant  on es   in   F ig ure   1 w hich   can  be  descr i bed   a s   fo ll ows   [ 11, 20] :   a)   Ca pacit [21] : t his r e pr ese nts  the am ou nt  of the sec ret m ess age th at   can  b e  carr ie d by the   cov e r fil e.   b)   Im per cepti bili ty   (steg file   qual it y):   it   con s iders  t he  qu al it of  the  ste go  aud i file   after   bein m erg e with the  secret   m essage.   c)   Secrecy   (Tem per   co nfr on ta ti o n ):  this  m eans  that  the  co ncea le m essage  is  secur e a nd   it   is  diff ic ult  f or  un a utho rized  pa rtie to  acce s it As   an   ex a m ple,  this  ca be  secret  key  that  is  s ha red  bet ween  th e   sen der  a nd the   receiver .   d)   Com pu ta ti on al   com plexity it   con ce rn s   th processi ng  tim ta ken   by   e m bed di ng  an ext racti on   op e rati ons,  w hich  s houl not  be  lo ng.  Especial ly this  app li es  to  const raine de vices,  s uch   as   sm artph on e .   e)   Robus tne ss:  this  enab le us   f or   rec overin the  secret  data  from   the  ste go  aud io  file   eve if  it   is  being  at ta cked .           Figure  1 .  Stega nograp hy m eas ur em ent p ar am et ers       In   a udio  data  hid in syst em s it   is  essenti al ly   req uire to  hav high  payl oad   w hile  m ain ta inin th e   qu al it y,  a nd  at   the  sam tim t he  e xistence   of   the  m essage  s hould  not  be  e asi ly   detect ed.   Th us m any  re searc pap e rs  a re  sti ll   lookin f orwa rd   t accom plish  these  iss ues This  is  because   HAS  ( H uma n   A udit or S yst e m is  an  obsta cl wh ic sta nds   against  data  hid in al gorit hm s   since  it   has  wide  dy na m ic   and   dif fere ntial    range  [22] .   Audio  data  hi ding  te ch nique us in Least   Sign i ficant  Bi (LSB)  s ubsti tuti on   te c hn i ques  [5]   ha ve   m uch   of  sim pl ic it a nd   popu la rity They  ca ha ve  great   pe rfor m ance  i te rm of  the   e m bed di ng  ca pacit and  com pu ta ti on al   c om plexity On  the  ot her  ha nd,  t heir  sec ur it le ve against  hac ker s   is  lo a nd  no t   com par able.  T he  sec ret  m essage  ca be  det ect ed  a nd  eve obta ine easi l y So m resear cher s   ha ve   so l ved  this  pro blem   by  com bin ing   LSB  ste gano gr a ph y   with  crypto gr a phic   schem in  order   to  i ncr ease  t he  se cur it y   le vel.  H ow e ve r,   this  is  no sat isfact or so luti on  du t the  wasted  com pu ta ti on al   com plexity   a nd   l os t   band width st or a ge.   In   this  pap e r,   we  pro pose  new   sc hem t hat  ex plo re di gital   op erat or s   (e.g.  X OR  NO gates )   m erg ed  by  an  LSB  substi tuti on   te c hn i qu e This  resea rch   i app li ed  i the   tem po ral  dom ai n,   the  au di cov e is   extracte int sam ples  and   are  conver te into  bin a ry.  I this  researc h,   eac sam ple  is   rep rese nted  by  16  bits.   In   the  em bed di ng   process,  the   po sit ions  of   t wo   ra ndom   bit are  sel ect ed  within  the  inte rv al   ( 5,   16)  as  key.   Accor ding  to  that  posit ion t wo   bits  from   e ach  sam ple  are  sel ect ed  to  be  us e as  g uid e   that  helps  rec overi ng  the  secret  m es sage.  T he  em b edd i ng   is  pe rfor m ed  by  app l yi ng   X OR  gate  on   the  secret   bits  and   the  sel ect ed   sam ple’s  bits.  The  f our  X OR  ou t pu ts  a re  c om par ed  with  t heir  c orrespo ndin sam ple’s  le ast   sign ific an bits.     If   they   are  not  m at ched   then  the  LSBs  are  change by  us i ng   a inv e rter  (NOT)   lo gical   op e rato r.   Ot he rw ise ,   there is  no c ha ng e . No rm al l y, this  helps  i ncrea sing t he dif ficult y of detect ing t he se cret m essage.     The  re st  of   t hi pa per   is  str uc ture as  the  f ollow i ng.  I S ect ion   2,  we   de scribe  s om e xisti ng   a udio   ste gano gr a ph schem es,  and   determ ine  their  streng t hs   an weaknesse s.  T he  detai of   the   pr op os e m e t hod  is   pro vid e in  Se ct ion   3.   Sect io n 4 prese nts t he  experim ental  res ults. T he  c oncl us i on s  are  draw i Sect io n 5.       2.   RE LAT ED  RE SEARCH   This  sect ion   pro vid es  t he  ba sic   theor ab out  so m aud io   data  hid in m et ho dolo gies   as  fo ll ows.   Audio  data  hi ding  resea rch   has  bee popula due  to  it s   stren gth for  exam ple  it si m pl ic it y.  Gen erall y,     aud i am pli tud values  a re  represe nted  in   rea nu m ber s,   so   durin the   e m bed ding  th secret  bits  do es  not  cause  m uch   noise   [23] It  is   norm al  that  e ach  schem has  it ow pro and   c ons.  A ll   are  m easur ed  an evaluate d usin the  af or em en ti on ed  d a ta  h i di ng  c rite rias.    Bi nn et   al [24]   has  pr opos e m et ho tha us es  LSB  of  each  sam ple  to  hid the  sec ret  inform at ion   wh il m ai ntain ing   the  ste go   file   i m per cepti bili ty Nev erthel ess,  they   fail   to   get  the  secrec and   the  r obust n es s   R o b u s t n e s s Q u a l i t y S e c u r i t y C a p a c i t y Evaluation Warning : The document was created with Spire.PDF for Python.
Ind on esi a J  E le c Eng &  Co m Sci     IS S N:  25 02 - 4752       Develo ping  audio   data  hid i ng sch e me usin g rand om s a m pl e b it s wi th     ( Mo hamm e d H atem Al i Al - H ooti )   149   and   c ould  not  fu ll rec ov e r   the  ori gin al   a ud i file In   t he  ne xt  re sear ch,   Al - Hooti  et   al [20]   des ign   a al gorithm   bas ed  on  m od ulus  f un ct io a nd   locat ion  m aps.   T his  af fects   the  qu al it and  the  rev e rsi bili ty ,     wh ic a re succ essfu ll ob ta in ed.   I c on tra st,  this  work is f r agile  to  sta nd a gainst  hac ker s  d et ect io n.   Chowdh ur et   al [ 25]   ha s   w orke on  a   m et ho base on  LSB   an it   is  c om bin ed   wit a   crypto gr a phic   schem called  patte rn   m at chi ng   al go rithm This  is  done   in   or de to  enc rypt  the  secret  te xt  so   it secrecy   is  gu a ran te e d.   By   usi ng   a   sim i la idea,  i [26]   th auth ors  pro pose  a al gorithm   that  i m ple m ents   LSB - based  dat hid i ng  in  a andr oid   e nviro nm ent.  This  sc hem sta rts  encr ypti ng  the  se cret  m essage,  a nd  put   it   in  MP3/WAV  au dio Me a nwhile the  c over  au dio   is  divi ded   int re gions.  Af te that,  the  secret  m es sage  is   e m bed de d by  m od ify ing  thes e re gions. H owever, the  em bed di ng cons um e s m uch  tim e to pro ces s.    Au t hors  of  [ 21 ]   has  c ombine Ge ner al iz ed  Diff e re nc Exp a ns io and   Re duce Dif fer e nce   Ex pan si on   to  pro du ce  ne w   schem that  i able  to  hid m or payl oad   and   m ai ntain  t he  qual it y.  Howev e r,    this  work   do e no pay  at te ntion   t the   secrecy Thi m et ho ha s   increase th capaci ty bu their  i m per cepti bili ty   m ay  n ot b e  hi gh  e noug h.   The  m et ho de scribe in  [ 27]   has  propose rev ers i ble  ste gano gr a phic   schem us ing   pro ba bili sti DNA - X OR  secret  sh ari ng - ba sed  col or   im ag e.  Si m il ar  to  ot her   res earc h,   e ach  D NA - X O truth   ta ble  va lue  is   evaluate by  usi ng   PS NR.  T he  input  that  ha the  hig he st  PSN val ue  is  con sid ere for  the  secret  sha rin g.  T hese  i nputs  a re  hi dd e i th co ver   im age  wh ic are   us e as  th key.   This  w ork  di vid es  the   secret  sh are s   into th ree  gro ups: R , G,  B wh ere the  secret  m essage is em bedde d.   As  it   is  m entio ne d,   we  f oc us  on   the  te m po ral  dom ai and   L SB  au dio  data  hid in re pr ese nts  hi gh  per ce ntage  fro m   the  cur re ntly   published   m et hods   in  this   do m ai n.   Mor eov e r,   L SB - ba sed  m et ho ds   s i m ply   increase  or  dec rease  one  valu fr om   the  LSB  value  of   eac sam pl [28] Howe ver,  thes schem es  are  no abl e   to  f ulfill   the   ot her   ste ga nograph ic   re qu i re m ents  su c as   secrecy   a nd  r obus t ness.  Als o,   t he  sec ret  m essage   m igh t be l os t e asi ly  if th e steg o fil e b ei ng att acked o c om pr esse [ 23] .       3.   RES EA R CH MENTH OD   This  sect io e xp li ci tl descr i bes  th pr opose al go rithm   of   this  resea rch.  Com m on ly LSB - base e m bed di ng  m e thods  ha ve  bee pro po s ed   in   m any  researc pa pers  [24] H ow e ve r,   t he  re searche rs   do  not   pay  m uch   at te ntion  to  t he  sec recy (Tem per   co nfr on ta ti on)   feat ure. A s a  re su lt , th os e   m et ho ds  are  le ss  us e fu l  in  the   current  pract ic al   data  hid ing  app li cat ion s This  is  because   of   the  secr et   data  extracti on   sim pl icity  an detect abili ty   [24] Currentl y,  secur e LSB  s chem es  are  only   the  on es  w hi ch  are  com bin ed  with  one  of   the  encr y ption   w orks  [ 25] This  c os ts  m uch   in  the  com pu ta ti onal   com plexity In   acc orda nce we  m ai nly  fo cus  on   increasin t he se cur it y pa ram et er.   Norm al l y,  com bin ing   data  hid in wit cr ypto  syst e m   [26]   can  inc rea se  the  secur it le vel  bu it   requires  m uch   com pu ta ti on al   com plexity Acco r ding  t thi co nd it io n,   w desig a L SB - bas ed  a udio  data   hid in m et ho by  co ns i de rin the  payl oad,  qual it y,  secrecy an com pu ta ti on al   com plexity   aspect s.    This  w ork  doe not  pay  at te nt ion   to   the  rob ust ness.   Dif fer e nt  from   the  pr e vious  re searc [22,   24] we  propose   new   te c hn i que  that  us e tw lo gical   op e ra tors  ( X OR  an NO T ba sed  on  LSB  s ub sti tu ti on T his  al go rith m   m ai nly  hid es  f our  bits  in  eac sam ple  w hich  is  pr e viously   co nv e rted   into   16  bits.  T hese   bits  are   di vid e in t two  par ts  as  i ndic at ed  in  Fig ure  2.  The  first  par is  use t carry  the  ou t pu ts  of   eac X O op e rati on,  a nd   t he   rand om  two  se le ct ed  bits a re  chosen  in  t he  s econd  pa rt that  contai ns   12  bits.           Figure  2 .   T he   us of au dio sa m ple r epr ese nted  in  b it s       3.1.      The  Proc edure  of Em beddin g   The  em bed ding  operati on  is  il lustrate as  sh ow in  Fig ure  and   as  sum m arized  in  t he  ne xt  fiv e   ste ps  as  foll ows:     Step  1:  read  the  sec ret  m essage  a nd  c onsider   each   f our   bits  ( , 1 ,   , 2 , , 3 , , 4 to   be   e m bed de within  eac h sa m ple.     Step  2:  rea t he   sam ples  of   a ud i file   (excl ud i ng  the  first  40  sam ples)  w her e   each   sam ple  is  in   th e   interval   [ 0,  65535],   an c onve rt  eac sam ple    into   16  bits.   Ever y   sam ple  is  div ide int t wo  par ts The   f irst  1 1 1 1 1 1 1 0 1 1 0 1 0 0 0 1 L S B M S B s e l e c t i o n   p a r t e m b e d d i n g   p a r t Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2502 - 4752   Ind on esi a J  E le c Eng &  Co m Sci,   Vo l.   13 , N o.   1 Ja nu a ry 20 19   :   147     154   150   on c onta ins  bits  w hich  a r us e to  be  c ha ng e d.   The  seco nd   pa rt  co ntain 12  bits  that  t wo   of  them   ,   a nd   ,   are   sel ect ed  ra ndom ly   to  be  proces sed   with   t he  first  a nd  t he   thir sec ret  bits  us i ng  XO R   ga te   (d e note in   this  pa pe by   ).  It  is  worth   to  note   that  t he  po sit ion   of   these   two  c hose bits  are   sent   to  t he   destinat io a a   key.      Step   3:   co ns i de eac f our   s ecret  bits  a nd  app ly   the   X O lo gical   op e r at or  as   s how in  (1).  Her e ,   , 1   is  the  outp ut  of  the  fi rst  X OR - in that  is  ap pl ie betwee t he  fir st  secret  bit  and   t he  fi rs sel ect ed  sam pl e   bin a ry   dig it   w her   is  an  in de that  re pr e se nts  the  num ber  of   the  us e sa m ples.  Both  the  two  values    and     are  in  the r an ge   of  5   an 16  w hic de note the  posit ion  o the  ra ndom   selected  bits  ,   an d   , resp ect ively .     It  is  com pu lso rily   that  bo th  r andom   po sit io val ues  sho uld   not  be  the  s a m e.  The  pa ra m et er  , 2   carries  the   ou t pu t   val ue  of  XO R - in the   seco nd  sec ret   bit  , 2   an t he  f irst  outp ut   , 1 To   increa se  the   c om plexity   of   com pr om isi ng   the  m et ho d,   t he  sec ond   rando m   sa m ple  bit  ,   is  X OR - e with  the  t hir secret  bit   , 3   Af te r  that,  we  i m ple m ent X O R t the  outp ut   , 3   with  fo ur th  se cret bit   , 4     , 1 =   , 1           ,     , 2 =   , 2           , 1     , 3 =   , 3           ,   , 4 =   , 4           , 3                                                                                                           (1)       Step  4:  em bed  al the  ou tp ut s   , 1 ,   , 2 ,   , 3 and   , 4   into  e ver sam ple  that  is  locat ed  in  the  first  par t.   Ba sed   on   ( 2) ,   the   outp ut are  c om par e with  eac fir st  par t   of  sam ple  bit   , 1 ,   , 2 ,   , 3 , , 4 If  the are  no t e qual , t hen each  sam ple b it  is flipp e d usi ng NOT l ogic al  o pe rato r.   Othe rw ise ,  t he re is  no c hange.     , 1 = { , 1 =  ( , 1 )                                     ,    , 1 =   , 1   , 1 = , 1                                                             ,        , 2 = { , 2 =  ( , 2 )                                     ,    , 2 =   , 2   , 2 = , 2                                                             ,        , 3 = { , 3 =  ( , 3 )                                     ,    , 3 =   , 3   , 3 = , 3                                                             ,        , 4 = { , 4 =  ( , 4 )                                     ,    , 4 =   , 4   , 4 = , 4                                                             ,        }                   (2 )       Step  5:  buil the  ste go  aud i file   based   on  the  new   m od if ie sam ple  val ues.   T hen   it   is  sent  to  the  receiver  v ia  any  p ubli c n et work.    As  a exam ple  for  this  em beddin pr ocedur e assum we  ha ve  a udio  sam ples  w hich  a re  conve rted   into  bin a ry  as  pr ese nted   in   Fi gure  4.  Th ese  t wo  sam ples  ar sub div ide i nt tw par ts nam el the  e m bed di ng   par wh ic is  r epr ese nted   by  four   le ast   si gn i ficant  b it s,  a nd   the  sec ond  pa rt  that  is  us e to  ge ner at t he   two  rand om l cho s en  bits.  S uppos the  sen der   c hoose the  bit  nu m ber   as  fir s po sit io an bit  nu m ber   10  as  the  seco nd   posit ion T he n,   base on  (1)  an (2)  the  ob ta i ne res ults  are   pr ese nted  i the  Tab le   w her ea s   su ccess fu ll the  ste go  f our   LSB  bits  are   rep la ced  by  the  f our  outp ut ( 1 , 2 , 3 , 4 of   the  E xclusi ve     OR ope rati on.           Figure  3 .  Th e  e m bed din g p r oc ess   T r a n s f o r m   i n   b i n a r y   f o r m m e s s a g e c o v e r O b t a i n   e a c h   f o u r   b i t s   t o   b e   e m b e d d e d C o n v e r t   i t   i n   b i n a r y   S e g m e n t   t h e   s a m p l e   i n t o   2   p a r t s G e n e r a t e   t w o   r a n d o m   v a l u e s     P r o c e s s   t h e   e m b e d d i n g   u s i n g   X O R   a n d   N O T   g a t e s B u i l d   t h e   s t e g o   a u d i o Evaluation Warning : The document was created with Spire.PDF for Python.
Ind on esi a J  E le c Eng &  Co m Sci     IS S N:  25 02 - 4752       Develo ping  audio   data  hid i ng sch e me usin g rand om s a m pl e b it s wi th     ( Mo hamm e d H atem Al i Al - H ooti )   151       Figure  4 .  Ex a m ple o the  em beddin g p ro ces s       Table  1 T he  E m bed din g E xa m ple O utp uts   A   b 1   O 1   b 2   O 2   b 3   B   O 3   b 4   O 4   1   0   1   1   0   0   1   1   0   1   1   1   0   0   0   0   1   1   0   1   0   0   0   1   1   1   0   1   0   1   1   1   0   0   0   0   0   0   1   1       3.2.      The  Ex tr act i on  Pr oced ure   It  is  com m on   that  the  extrac ti on   pr ocess  is   j us i nv e rsing   the  em bed ding  ope rati on.  T he  recei ver  ta kes  the  deliv ered   ste go  file   an tw c hos en  key  val ues   that  in dicat es  the  posit io ns   of   t he  tw sel ect ed  sam ple b it s.  Aft er th at , t his  procedu re is  perf or m ed  as s how in  Fig ure  a nd the  f ollow i ng ste ps :     Step  1:  ext ract  the  ste go  au di file   into  s am ples  wh ic are   co nv e rted   into   bin a ry.  C orres pond  to   th e   e m bed di ng pr oc ess, the  f ir st 4 sam ples are  i gnored .     Step  2:  inse rt  the  tw key  va lues  ,     into  the  extracti on  syst e m   to  determ i ne  the  po sit io ns  of   th e   rand om l y sel ect ed  sam ple two bits   , ,   , .     Step  3:  a pp ly   XO R   oper at or  betwee t he  fir st  sel ect ed  sam ple  bit  ,   a nd  the   first   LSB  of  t he  ste go  sam ple.  As  a   r esults,  t he   first   secret   bit   , 1   is  obta ined This   f ir st  secret   bit  is  al so  XO R - ed   with   the   se c on LSB  of   t he  ste go   sam ple  in  order   t obta in   the  seco nd   se cret  bit   , 2 A fter  that,  the  thir secret  bit  , 3   is   ob ta ine d by X OR - in the   thi r LSB   of  t he   s a m sam ple  , 3   a nd  the   sec ond sel ect ed  rand om   b inary d igit   ,   Finall y,  getti ng  the  f ourth  sec r et   bin ary  dig it   , 4   XO is  a pp li e to  th obta in ed  thir sec ret  dig it   , 3   an th e   sam ple f ourth LSB   , 4     as   in E qu at ion   (3).           Figure  5 .  T he  e xtracti on  op e ra ti on       , 1 =   , 1           ,     , 2 =   , 1           , 2     , 3 = , 3           ,   , 4 =   , 3           , 4       (3)     H i 1 1 1 1 1 1 1 0 1 1 0 1 1 1 0 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 1 0 1 1 0 1 0 0 0 L S B M S B S e c r e t   m e s s a g e 0 1 1 0 1 0 0 1 6 5 2 3 5 6 5 4 0 1 4 1 5 7 2 6 3 3 6 0 A u d i o   s a m p l e s A u d i o   s a m p l e s   i n   b i n a r y 0 1 0 0 1 0 0 0 O b t a i n   t h e   s e c r e t   m e s s a g e   a n d   t r a n s f o r m   i t   i n t o   i t s   o r i g i n a l   f o r m m e s s a g e S t e g o C o n v e r t   i t   i n   b i n a r y   S e g m e n t   t h e   s a m p l e   i n t o   2   p a r t s P r o c e s s   t h e   e x t r a c t i o n   o p e r a t i o n I n s e r t   t h e   t w o   r a n d o m   p o s i t i o n s Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2502 - 4752   Ind on esi a J  E le c Eng &  Co m Sci,   Vo l.   13 , N o.   1 Ja nu a ry 20 19   :   147     154   152     Step  4:   finall y,   com bin e   al of   the   ob ta ine s ecret  bits  , 1 , , 2 , , 3 ,   and  , 4   wh ic a re   obta ined   f ro m   eac sam ple in  a v ect or an c onve rt that  vec tor  i nto  it or i gi nal form  that m igh t be te xt, a ud i o,   or im age.       4.   RESU LT S   A ND AN ALYSIS   In   t his  re searc h,   t he  sec ret  m essage  is  ge ner at e us in Ra nd functi on   in  (Mat la b - 2017 a ).   For  t he   aud i co ver,  we  us 10   a udio  * .w a file s.  Each  co ve r   file   has  it own  du rati on  and   siz e,  ho wev e r,    al are  sa m ple in  16   bit  dept [29 - 31] It  i intende that  the  file are  c ho s en  in  dif fe r ent  cases  e.g s peech,   m us ic So m a re  in  high  le vel   and  ot her has   low  s ound  le ve l.  The  nu m ber  of   sam ples  (excl ud in t he  40  on e wh ic ar ig nored   since  t hey  can  be  im pacte d)   is  va ried.  T his  sim ula ti on   is  done  i c om pu te la ptop  Asus   that  has  co re  i3  proces sor,  GB  RAM,  an 500  GB  H D Four   sce nar i os   are  c onduct ed  to  eval uate  fou r   featur e s as  f ollow s:   First, ca pacit y or  em beddin r at e that i s r ep re sented  a s s how in  Table  2.   T his m et ho is a ble to   hid e   secret  bits  ti m es  the  num ber   of  t he  co ve sam ples  wh ic m eans  each  sa m ple  is  able  to  carry  bits  w her ei the  existi ng  w orks  [20,  24]   e ach  sam ple  can  on ly   car ry  one  bit H ow e ve r,   the   m or th inserte da ta   causes   sli gh t i m pact o the  quali ty  o f  the ste go f il e a s in dicat ed  in  both Ta ble  a nd Ta ble  4.     Seco nd,  im per cepti bili ty   wh i ch  is  m easur e by  us in Si gnal   to  N oise  Ra t io  (SNR)   an Peak  Sig nal  to  No ise   Ra ti (PSNR as   in   E quat io (4)  and  ( 5),  res pe ct ively   [8] Th va riable    de no te s   the   num ber  of   sam ples,  an m eans   the   c over  a udio   w he re   the   ste go  file   is  pr ese nted   by     wh il e   µ   re pr ese nt the   m axi m um  value  of  au dio   sam ple .   The  hig he the  value  of   S NR  and   P SN R,  the  great er  the  si m il arity  between   the  cov e and the  ste go  fi le s.      = 2 = 1 ( ) 2 = 1   (4)      = µ 2 ( ) 2 = 1   (5)     Thir d,   the  sec ur it par am et er  is  i m pr ove in  this  pa per.  This  is  beca use   in  m os exist ing   m od ul us   and  LSB  w ork su c as   [ 20,  24 ] t he  sec ret  m essage  rec ov ery  is  sim ply  ob ta ined   by  ta kin the  LSB  of   each   ste go   sam ple.  I co ntrast,  t his w ork  hi des  the   outp uts  that are  obta ine f r om   XO R - in ga te in  each  sa m ple.   These  outp uts   are  not  th ex a ct   secret  bits.  Ther e f or e,   s uppo s the  at ta c ke kn ows  our  m et ho dolo gy,  he/she   needs  to  c ondu ct   16   t rial for  each  sam ple  in  orde to obtai the  e xact 4   s ecret  bits.   In   t hi pap e r,  w e d e sign e an  e qu at io t ha is  pr ese nted   in  ( 6).  It  is  us e d   to   m easur the  am ou nt  of  tria ls  that  are  re qu i red   f or   obta inin the  sec ret  m ess age.  He re,    is  t he  nu m ber   of  t i m es  need e t il le gally   ob ta in  the   sec ret  bit s,    re pr ese nts  t he  four  sel ect ed  bi ts,    is  the  pos sibil it ie of   ea ch  sec ret  bit  ( since  it   is  i bi nar s   is  al ways  2),   an   is  sy m bo li zed as t he  num ber  o f t he  ste go sam ples that are em bedde d.   As  a r esult, h ac kers  need   16 trial s for  eac sam ple  in  or de to  obta in  the  exact  f our  secret  bits.  Th is  is  m ul ti plied   by  the  num ber   of   the  em bedd e sam pl es. Th ere fore,  the m or the am ou nt of  hidden sec ret bi ts t he  higher  t he  di ff ic ulty  fo hac kers to  de te ct  it.   Ba sed  on  E qu a ti on   (6),  le the   em bed de ste go  sam ples  am ount  is  6615 00  an it   is  know in   eac sam ple  4 2   trai ls  are  nee de in  orde to   ob ta in   the  4   se cret  bits,  s t he   possibil it of  il le gally   gaining  the  whole  s ecret  m essage  is   ( 4 2 ) 661500 = 1 0 5 8 4 0 00   ti m es,  wh e reas  in  the  e xisti ng  wor ks   [ 20,  24]   al low  a nyone  t directl y extract  the sec ret m es sage  by  ob ta ini ng the  LSB  of  each sam ple.      = ( 2 )   (6)       Table  2 . T he  E m bed din g R at base d o F our  Scen a rios   No   File na m e   No Sa m p l es   1 0 0 % p ay lo ad   7 5 % p ay lo ad   5 0 % p ay lo ad   2 5 % p ay l o ad   1   0 0 1 _ 0 1   7 5 2 0 6 8   3 0 0 8 2 7 2   2 2 5 6 2 0 4   1 5 0 4 1 3 6   7 5 2 0 6 8   2   0 0 1 _ 0 2   8 1 5 5 2 0   3 2 6 2 0 8 0   2 4 4 6 5 6 0   1 6 3 1 0 4 0   8 1 5 5 2 0   3   AM2_ 3 5 U_ 5   5 7 8 3 3 7 7   2 3 1 3 3 5 0 8   1 7 3 5 0 1 3 1   1 1 5 6 6 7 5 4   5 7 8 3 3 7 7   4   Aco m ic   6 6 1 5 0 0   2 6 4 6 0 0 0   1 9 8 4 5 0 0   1 3 2 3 0 0 0   6 6 1 5 0 0   5   0 0 1 _ 0 3   9 9 7 1 1 9   3 9 8 8 4 7 6   2 9 9 1 3 5 7   1 9 9 4 2 3 8   9 9 7 1 1 9   6   Blu es   6 6 1 5 0 0   2 6 4 6 0 0 0   1 9 8 4 5 0 0   1 3 2 3 0 0 0   6 6 1 5 0 0   7   Clas sical   6 6 1 5 0 0   2 6 4 6 0 0 0   1 9 8 4 5 0 0   1 3 2 3 0 0 0   6 6 1 5 0 0   8   0 2 .Scho o l Bo y - 9   1 7 6 3 9 9 9   7 0 5 5 9 9 6   5 2 9 1 9 9 7   3 5 2 7 9 9 8   1 7 6 3 9 9 9   9   Jazz   6 6 1 5 0 0   2 6 4 6 0 0 0   1 9 8 4 5 0 0   1 3 2 3 0 0 0   6 6 1 5 0 0   10   Ro ck   6 6 1 5 0 0   2 6 4 6 0 0 0   1 9 8 4 5 0 0   1 3 2 3 0 0 0   6 6 1 5 0 0       Evaluation Warning : The document was created with Spire.PDF for Python.
Ind on esi a J  E le c Eng &  Co m Sci     IS S N:  25 02 - 4752       Develo ping  audio   data  hid i ng sch e me usin g rand om s a m pl e b it s wi th     ( Mo hamm e d H atem Al i Al - H ooti )   153   Table  3 . T he  S te go   A ud i o Quai lt y based on  Sign al  t o No is e Rat io   No   File na m e   SNR v alu es in  dB   In1 0 0 %   In7 5 %   In5 0 %   In2 5 %   1   0 0 1 _ 0 1   7 7 0 2   7 8 .27   8 0 .03   8 3 .02   2   0 0 1 _ 0 2   7 6 .97   7 8 .25   8 0 .02   8 3 .03   3   AM2_ 3 5 U_ 5   7 6 .62   7 7 .86   7 9 .61   8 2 .58   4   Aco m ic   7 4 .46   7 5 .71   7 7 .47   8 0 .48   5   0 0 1 _ 0 3   7 7 .02   7 8 .28   8 0 .05   8 3 .07   6   Blu es   7 4 .05   7 5 .30   7 7 .06   8 0 .08   7   Clas sical   7 4 .92   7 6 .17   7 7 .92   8 0 .94   8   0 2 .Scho o l Bo y - 9   7 6 .48   7 7 .72   7 9 .48   8 2 .49   9   Jazz   7 5 .94   7 7 .19   7 8 .96   8 1 .97   10   Ro ck   7 6 .10   7 7 .35   7 9 .11   8 2 .12       Table  4 . T he  S te go   A ud i o Quai lt y based on  Peak Si gnal  to Noise  Rat io   No   File na m e   PSNR in  dB   In1 0 0 %   In7 5 %   In5 0 %   In2 5 %   1   0 0 1 _ 0 1   8 0 .04   8 1 .29   8 3 .04   8 6 .04   2   0 0 1 _ 0 2   7 9 .98   8 1 .26   8 3 .03   8 6 .04   3   AM2_ 3 5 U_ 5   8 0 .02   8 1 .27   8 3 .01   8 5 .99   4   Aco m ic   8 0 .05   8 1 .30   8 3 .06   8 6 .07   5   0 0 1 _ 0 3   8 0 .02   8 1 .28   8 3 .05   8 6 .07   6   Blu es   8 0 .04   8 1 .29   8 3 .04   8 6 .06   7   Clas sical   8 0 .05   8 1 .30   8 3 .06   8 6 .08   8   0 2 .Scho o l Bo y - 9   8 0 .05   8 1 .29   8 3 .05   8 6 .06   9   Jazz   8 0 .0 5   8 1 .30   8 3 .06   8 6 .08   10   Ro ck   8 0 .06   8 1 .31   8 3 .06   8 6 .07       Four t h,   c om pu ta ti on al   c omplexit y,  this  m et ho c os ts   in  av era ge  near ly   30. 20  seco nd s   to  be   i m ple m ented.   This  fast  com puta ti on al   proce ss  al lows  us   to   instal the  pro po s ed   syst em   even   in   sm artph ones .   B ased  on  the  c orrelat ion  m et ho d   [ 20]   the  fu ll  r ec ov e ry of t he  secret m essage is   100%   gua ran te e d .   In   br ie f this  work   has  s uc cessf ully   ob ta ined  gr eat   cap aci ty   and   m ain ta ined  ste go   fil qu al it y.     The  sec recy  is   al so   high  in  a   fast  com pu ta ti on al   c om plexity   perform ance.  In  dif fere nt  w ords,  this  al go r it h m   can  e nh a nce   th capaci ty sec recy  an c om plexity   wh il m ai ntainin the  q ualit y.  O t he   oth e ha nd,  thi w ork   is  no rob us to  sta nd  agai ns any  at ta cks  e.g WAV  to  MP3  com pr ess ion T heref or e ,   any  at ta ck  on  it   ca n   cause l os in th e secret m essage  wh ic ca n b e co ns tr uctive i s om e cases.       5.   CONCL US I O N   Audio  data  hidi ng   has  it secur it sign ific a nc in  m any  field su c as  bankin g,   m i li ta ry,  and   ai rlines   syst e m s.  In   t hi pa per,  we  present  s chem that  f oc us es   o n   im pr ov in the   ca pacit y,  com plexity an the  secrecy   crit eri of   a L SB  su bst it ution   al gorithm   without  bei ng   c om bin ed  with  c r yptogra ph sc hem e .     The  pro ba bili t of   detect ion  is  m ini m iz e d   by  us ing   X O gates  wh er eas  the  com pu ta ti on al   com plexity   perform ance  i achieve d.   I the  w or st  case,  this  sch e m con su m es  around  30. 20   sec onds   wh e rei crypto gr a phic   schem es  can  consum m uc of   c om pu ta ti on al   com plexity   and   wast m uch   of   st or a ge.    As  com par iso bet ween   our   m et ho an norm al   LBS  wo r ks   we  a re  a ble  to  achie ve   m or em bed di ng   rate  wh il m ai ntaining   the  qual it y.  More over th is  m et ho ca yi el gr eat   secrecy   com par ed  t the  e xis ti ng   relat ed  sc hem e s.  I the  f uture we   ar plan nin to   inc rease  the  em bed di ng  rate  s each   s a m ple  can  car r one   byte W belie ve  this  will   not  on ly   increase payl oad   but  al so   it   will  fu r ther  com plica t the  secret  m essage   detect ion.       ACKN OWLE DGE MENTS     The  a uthors  w ou l li ke  to  e xpress  t heir  warm est   gr at it utes  to  Insti tut  T ek no l og Se pulu Nopem ber   ITS, S ur a baya,  Indo nesia f or the s upport t hat h as  b ee n give to  this  resear ch.         REFERE NCE   [1]   Bhui y an  SS N,  et   al.  An  Im prove Im age   Ste ganogr aph y   Alg orit hm   base on  PV D.  Indone sian  Journal  of   El e ct ri ca l   Eng in ee ring   and   Com pute Sc ie nc e. 2 018;10: 569~77.   [2]   Din  R,   et   al.  Com par at ive  Re vie on  Da ta   Hi ding  Schemes.  I ndonesia Journ al   of   Elec tr ical  Engi ne eri ng  an d   Com pute Scie n ce .   2018;11 :768 ~74.   Evaluation Warning : The document was created with Spire.PDF for Python.
                          IS S N :   2502 - 4752   Ind on esi a J  E le c Eng &  Co m Sci,   Vo l.   13 , N o.   1 Ja nu a ry 20 19   :   147     154   154   [3]   Al - Dm our  H,  A l - Ani  A.  steg anogr aph y   embedding  m et hod  b ase on  edg id ent ifica ti on  and   XO codi ng .   Expe rt   s y st ems   with  Appli ca t ion s.  2016;46: 293 - 3 06.   [4]   Ahm ad  T,   Fiqar  TP.   Enha nc in the   Perform anc of  Audio  Data   Hiding  Method  b y   Sm oot hing  Inte rpol at e d   Sam ple s.  Int ern a ti onal Journal of Innovati v Com puti ng,   In form ation  and   Control. 2018;14(3):757 - 79.   [5]   Begum  MB,  Venka t ara m ani   Y .   LSB  base d   audi steg anogr aph y   base on   te xt   compress ion.   Proc edi a   Engi ne eri ng.   20 12;30: 703 - 10.   [6]   Hus sain  HS ,   et   al .   The   Embeddin Perform anc of  StegSVM   Mod el   in  Im age   Steg anogr aph y .   Indo nesia Journal   of  Elec tr ical Eng ine er ing  and   Co m pute Scie n ce. 2018;12:233~8.   [7]   Al  Huti  MH A,  et   al.,   ed it ors.  In c rea sing  th ca p a ci t y   of  th sec r et  dat using  DEpi xel bloc ks  and  adj usted  RDE - base on  gra y sc al images .   Inform at ion  Com m unic at ion  T echnolog y   and  S ystems   (ICTS),   2 015  Inte rna t ional   Confer ence  on ; 2015:  IE EE.   [8]   Mansor  FZ,   et   al .   An  Anton y m   Subs ti tut ion - base Model  on  Li nguisti Ste ganogr aph y   Me t hod.   Indone sian   Journal  of El ec tr ic a Eng ineeri ng   and  Com puter S ci en ce.  2018;12 : 225~32.   [9]   Devi  RR,  Pugaz hent hi  D.  Ide al   Sam pli ng  Rat t Reduc Distor ti on  in  Audio  Stega nogra ph y .   Proce di Com pute Scie nc e. 2016; 8 5:418 - 24.   [10]   Rahmani  P,  Dastgha ib y far G.  An  eff ic i ent   histog ram - base inde x   m appi ng  m ec hani sm   for  rev ersibl data  hidi ng  in  VQ - com pre ss ed  images .   Infor m at ion  Sci ences.   2017.   [11]   Shahadi   HI,  e a l.   Concurr ent   ha rdware   ar chi t ec t ure   for  dual - m o de  audi st ega n ogra ph y   p roc ess or - base FP GA .   Com pute rs &   E l ec tr ic a l Engi ne er ing.   2016 ;49: 95 - 116.   [12]   Sloan  T,  Hern ande z - C astro  J.   Forensic  an al y sis  of  v ide stega nogra ph y   t ools.  Peer Co m pute Scie n ce.   2015;1: e7 .   [13]   Yan  F,  et  a l.   F le xible  r epr ese n ta ti on   and  m anipulat ion   of  aud i signal on  qu ant um   computer s.  The or et i cal   Com pute Scie n ce .   2017.   [14]   Hu  H - T,   et   al.   S y nchr onous   b li nd  a udio  wa term ark ing  via  shape   conf igur at i on  of  sorted   L W coe ff ic i ent  m agni tude s.  Sig nal   Proc essing.   2 018;147: 190 - 20 2.   [15]   Hua  G,  et   al.  Twe nt y   y ears   of  digi tal  audi o   wate rm ark ing comprehe n sive  rev ie w .   Signal   Proce ss ing .   2016;128: 222 - 4 2.   [16]   Le BY,  et   al .   Bl ind  and  ro bust  audi wat ermarki ng  sche m base on  SV D DCT.   Sig nal   Proce ss in g .   2011;91(8): 1973 - 84.   [17]   Kar  DC,  Mulkey   CJ.  m ult i - th reshold  base a udio  stega nogra ph y   sche m e.   Jou rna of  information  sec urity   and   appl i ca t ions.  20 15;23: 54 - 67.   [18]   Renz D ,   L em us   C.   Authen ti c ity  ver ifica ti on  of  a udio  signal bas ed  on  fra g il w a te rm ark ing  for  a udio  fore nsics .   Expe rt   S y st ems   with  Appli ca t ion s.  2018;91: 211 - 2 2.   [19]   Chang  C - C.   r eve rsibl data  h idi ng  sche m us ing  complement ar y   embedding  strat eg y .   Inform at ion  Sci enc es.   2010;180(16): 30 45 - 58.   [20]   Al - Hooti  A,  et   a l.   Audio  Dat Hi ding  Based  on  Sam ple   Value   Mo difi c at ion  Us ing  Modulus  Functi on.   Journal  of   Inform at ion  Pro ce ss ing  S y st ems .   2016;12(3).   [21]   Andra  MB,  et   al .   Med ical  Re cor Protection   with  Im prove GRD Data   Hi ding  Method   on  Audio  Files .   Engi ne eri ng  Lett ers.   2017;25 (2).   [22]   Li   X,  Yu  HH ,   edi tors.   Tra nsp ar ent   and  robust  a udio  dat hid ing   in  subband  dom ai n.   Inform at i on  Te chnol og y :   Coding  and   Co m puti ng,   2000   Proce ed ings I nt er nat ion al   Conf erence   on 2000 IE EE .   [23]   Datt a   B,   et  al.,  edi to rs.  Mult i - bit   Dat Hid ing   in  Randoml y   Chosen  LSB  L a y ers  of  an  Au dio.   Inform ation  Te chno log y   (ICI T),   2016   Int ern a t iona l   Confer e nc on;   2016:   IEEE.   [24]   Binn y   A,  Koilakuntla   M,  edi to rs.  Hiding  sec ret   informati on  using  LSB  bas ed  audi s te ga nogra ph y .   Soft   Com puti ng  and M ac hine Int el l ig enc e   (ISCM I),   2 014  Inte rn at ion a Confer ence  on;   2014:  I EEE.   [25]   Chowdhur y   R ,   et   al .   vi ew  o LSB  base d   a udio  steg anogr a ph y .   Int ern ation al   Journa of  S ec uri t y   and   Its  Applic a ti ons.  20 16;10(2): 51 - 62.   [26]   Si buria R,   edit or  Stega nogra ph y   implement at io on  andr oid  smart phone  using  t he  LSB  (le ast  si gnifi c ant   bit to   MP and  W AV   audi o.   W ire l ess  and  T el ema ti cs  ( ICW T),   2017  3r Inte rn at ion al C onfe ren c on 20 17:  IE EE.   [27]   Tunc er   T ,   Avci   E.   r eve rsib le   d at a   hidi n g   al gor i thm  base on  pr obabi li st ic  DN A - XO sec ret   sha ring  sche m fo r   col or im age s.  Di spla y s .   2016;41 : 1 - 8.   [28]   Basu  PN ,   Bho wm ik  T,   edi tors .   On  embeddin of  t ext   in  au dio  a   c ase   of   stega nogr aph y .   Rec en Tr ends  i n   Inform at ion, T e l ec om m unic at ion   and  Com puti ng   (IT C),   2010   Int e rna ti on al   Conf er enc e   on;   2010: I EE E .   [29]   Tz an et ak is G. Dat ase ts  -   m ars y as   [onli n e] .   2002.   [30]   Urbana - Champa ign  SS TGUoIa.  Ill ino is  Speec h   and  L angua g Engi ne eri ng.   Univer sit y   of   Illi nois  at   Urban a - Champaign;   200 3.   [31]   Bosch  JJ ,   et   al .   P.   Co m par ison  of  Sound  Segre gation  Te chn iq ues  for  Predom i nant   Instrum ent   Rec ognition  in   Mus ic al   Audio   Signal s.   IRMA S: a   da ta set   for   inst rum ent   re cogni t i on  in  m usical  au dio  signa ls.      Evaluation Warning : The document was created with Spire.PDF for Python.