APost-ProcessingAlgorithmAppliedtoReduce
theSizeoftheCoveringArraysoftheNIST
Repository
Ph.D.Jose
Torres-Jimenez
LaboratoriodeTecnologıasdeInform
acion,CINVESTAV-Tamaulipaskm
5.5
Carretera
Cd.Victoria-Soto
laMarina,87130,Cd.VictoriaTamps.,Mexico.
September
19,2013
1/72
Contents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
2/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
3/72
Introduction
SoftwareTesting
�Now
adayssoftwareproductsarepresentin
almostallhuman
activities,then
itisnecessary
toassure
ahighlevelof
function-
alityof
thesoftwareproducts.
�Itisestimated
that
atleast50%
ofthecost
ofdevelopinga
new
softwarecomponentisrelatedto
thetestingprocess.1
1B.HailpernyP.Santhanam
,2002[1].
A.Hartm
an,2005[2].
4/72
Introduction
SoftwareTesting
�Oneoption
totestasoftwarecomponentisto
usean
exhaustive
approach,i.e.
test
allthepossible
combinationsof
theinput
param
eters.
�Another
option
isto
use
combinatorialtesting,
that
guarentees
that
allthecombinationsof
certainnumber
ofparam
etersis
tested
exactlyor
atleastcertainnumber
oftimes.2
2Cohen
etal.[3]yCohen
etal.[4].
5/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
6/72
BasicDefinitions
Orthogonal
Arrays(OA)
Definition
Let
beN,t,
k,v,andλfive
positive
integers,an
orthogonalarray
OAλ(N
;t,k
,v)isan
N×karrayA=(a
i,j),0≤
i≤
N−1,
0≤
j≤
k−1,
over
Zv=
{0,1,...,v
−1}
withtheproperty
that
foranytdistinct
columns,thereareexactlyλrowsthat
takeseach
valueof
Zt v=
{0,1,...,v
−1}
t.
�When
λ=1theOAisof
unitaryindex.
�For
OAsof
unitaryindex
andvaprimepow
erthereexists
anoptimalsolution
3
3Bush
[5]
7/72
00
00
00
01
00
11
01
01
01
01
10
01
01
11
10
11
01
01
10
01
10
01
11
10
01
11
00
01
Figure
1:TheorthogonalarrayOA2(8;2,7,2).
8/72
BasicDefinitions
Orthogonal
Arrays(OA)
�Thecase
when
visaprimepow
erandv≥
t,issolved
using
arithmeticof
aGaloisFiniteField.Thevalueof
each
cellof
theOAistheevaluationof
thenumber
ofrowin
thenumber
ofcolumn,thelast
columnisdirectlytheleftmostcoeffi
cient
ofthenumber
ofrow.Thisprocedure
canbeimplemented
usingalogaritm
tableforaGaloisFiniteField
4
�Thecase
when
visaprimepow
erandv<
tissolved
using
thezerosum
algorithm.
4Torreset
al.[6]andTorreset
al.[7]
9/72
BasicDefinitions
CoveringArrays(CA)
TheOAsareagoodoption
totesteffi
cientlysoftwarecomponents,
butthey
havethedrawbackthat
they
donot
existforcertaincom-
binationsof
param
eters(λ,k
,v,t).
Arelaxedversionof
theOAs
aretheCAs(eachcombinationmust
existat
leas
once).
Given
the
relaxation,they
existforanycombinationof
param
etersandthey
havealower
number
ofrows.
Definition
Let
beN,t,
k,andvfourpositiveintegers,acoveringarray
CA(N
;t,k
,v)5
isan
N×karrayA=(a
i,j),0≤
i≤
N−1,
0≤
j≤
k−1,
over
Zv=
{0,1,...,v
−1}
withtheproperty
that
foranytdistinct
columns,at
leastonerowtakeseach
valueof
Zt v=
{0,1,...,v
−1}
t.
5when
thecolumnshavedifferentorder
itiscalledaMixed
CoveringArray
(MCA)
10/72
BasicDefinitions
CoveringArray
Number(C
AN)
Thecoveringarrayconstructionproblem
(CACP)consistsin
con-
structingacoveringarrayCA(N
;t,k
,v)giventheparam
eterst,
k,
andvin
such
away
thenumber
ofrowsN
ofthecoveringarray
isminimal.Thesm
allest
Nforwhichacoveringarrayexists
isthe
coveringarraynumber
(CAN)fortheparam
eterst,
k,andv,and
itisdenoted
by CAN(t,k
,v)=min{N
|∃CA(N
;t,k
,v)}.
(1)
11/72
BasicDefinitions
NP-Complete
andSearchSpaceforCAs
�Even,thereisnoproof
that
CACPbelongs
totheclassof
NP-
Complete
problems,somerelatedproblemsareNP-Com
plete.
For
instance:theproof
ifitispossibleto
addanew
rowthat
provides
atleastacertainnumber
oft-wisemissingcombina-
tions6
�Thesearch
spaceof
theCACPdenoted
bySsatisfies:
vt≤
S≤
�vk N
�
(2)
6Colbourn
[8]
12/72
BasicDefinitions
Optimal
CoveringArrays
Therearereportedonlyasm
allnumber
ofoptimalcoveringarrays:
�Thecase
CA(v
t;t,m
ax(v,t)+1,v)when
visaprimepow
er7.TheCAsconstructed
inthisway
areequivalentto
OAsof
unitaryindex.
�Thecase
(v=2)
∧(t
=2)
wherek≤
� N−1
�N 2�
�8
Ingeneralthedeterminationof
theCANisvery
difficultandisthe
them
eofalotof
research,butasym
ptotically
(for
largek)
CAN(t,k
,v)≈
vtlog(k)
(3)
7Bush
[5]
8Renyi
[9],Katona[10],Kleitman
andSpencer[11]
13/72
BasicDefinitions
Isomorphism
inCAs
Given
that:
�Thepositionof
therowsin
aCAisnot
relevant,alltheCAs
obtained
bypermutingtherowsof
aCAareisom
orphic.
�Ifallthecolumnsin
aCA
havethesameorder
(alphabet),
alltheCAsobtained
bypermutingthecolumnsof
aCA
are
isomorphic.
�Thecoverage
properties
ofaCAarenot
affectedifthesym-
bolsin
onecolumnarepermuted(for
instance
allzerosare
exchangedwithallones),
alltheCAsobtained
bypermuting
symbolswithin
columnsareisom
orphic.
Definition
For
aCAthereareN!k!(v!)kisom
orphicCAs.
N!permutation
ofrows,k!
permutationsof
columns,and(v!)kpermutationsof
symbolsin
thecolumns.
14/72
Given
twoCAsAandBCA(6;2,5,2)they
areisom
orphicifonecan
beconstructed
from
theotherusingsomepermutation
ofrows,per-
muationof
columns,andpermutation
ofsymbols.
Consideringthat
τdefines
therow
permutation,πisthecolumnpermutation,and
φdefines
thesymbol
permutation.In
thenextfigure
weillustrate
how
toconstruct
Bfrom
A.
A=
11
10
11
01
10
00
00
11
10
10
01
10
00
11
11
B=
01
10
10
01
00
11
01
01
00
01
00
01
11
01
11
τ=(0
12345)
π=(0
1234)
φ=(0
0000)
τ�=(3
12540)
π�=(4
2031)
φ�=(0
1010)
Figure
2:UsingCAAweobtain
CAB
usingτ� ,π� ,andφ� .
15/72
BasicDefinitions
Redundancy
inCAs
ThetwocasesofoptimalCAsmentioned
previosuly(O
Aswith(λ
=1)∧(primepow
er(v)),and(v
=2)
∧(t
=2))resultin
CAsthat
do
not
haveredundancy.For
anyother
case
thepossibility
that
aCA
has
alotof
redundancy
isvery
high.Aredundantelem
ent(usually
denotedas
wildcard)cantake
anyvalueandthecoverage
properties
ofaCAarenot
affected(usually
arerepresentedwiththesymbol
*).
Definition
Oneelem
entof
aCAisredundant(wildcard)ifwecanchange
that
elem
entwithanyvalueof
itsalphabet
andthecoverage
properties
oftheCAarenot
affected.
16/72
For
instance
theCAACA(11;3,5,2)
donwloaded
from
theNIST
Repository
9has
onerowtotally
redundant.
TheCAB
indicates
with*theredundantelem
ents.
A=
00
01
00
01
01
01
00
00
11
10
01
01
11
10
11
11
10
11
00
01
11
01
01
01
11
10
10
0
B=
00
01
00
01
01
01
00
00
11
10
01
01
1∗
∗∗
∗∗
11
10
11
00
01
11
01
01
01
11
10
10
0
Figure
3:CAAcanbereducedin
onerowthrough
thedetection
of
redundantsymbolsshow
nin
CAB
as*
9http://math.nist.gov/coveringarrays/ipof/tables/table.3.2.htm
l17/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
18/72
CARepositories
CA
Repositories
Therearethreemainrepositoriesforuniform
CAs:
�NIST
Repository
10that
containsexplicitCAsconstructed
usingthe
IPOG-F
algorithm
for(v
={2,...,6})∧(t
={2,...,6}).
�CharlesColbourn
Repository
11that
lists
only
thebestknow
nsizesfor
CAs(v
={2,...,25})∧(t
={2,...,6},itdoestnotprovideexplicit
CAs.
�CinvestavCA
Repository
12currentlyitprovides
explicitCAs.
Itcontains
goodCAsfor(v
={2,3})∧(t
={2,...,6}).
10http://math.nist.gov/coveringarrays/
11http://www.public.asu.edu/∼ccolbou/src/tabby/catable.htm
l12http://www.tam
ps.cinvestav.m
x/∼jtj/authentication.php
19/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
27/72
ConstructionMethods
Dueto
thediffi
cultyofsolvingtheCACP
anumber
ofmethodshavebeen
developed.Themethodscanbeclassified
infive
maincategories:
�Exact
methods.
Thesemethodsguaranteeto
findan
optimalCA,butgiven
theexponential
search
spacethey
arepractical
only
forsm
allCAs.
�Greedymethods.
Thesemethodsdonotguaranteeto
findan
optimal
CA,
butthey
arevery
fast
andcanbeusedto
construct
anyCA.
�Metaheuristic
methods.
Thesemethodsdoes
notguaranteeto
findand
optimal
CA,they
take
moretimethan
thegreedymethodsandin
many
casesthey
givebetterCAsthan
theones
obtained
usinggreedymethods.
�Algebraicmethods.
Thealgebraicmethodsinvolved
form
ulasor
operations
withmathem
atical
objectssuch
asvectors,finitefields,
groupsor
another
(usually)sm
allcoveringarrays.
�Manipulationmethods.
Thesemethodsuse
coveringarrays
previouslycon-
structed
toconstruct
new
ones,or
transform
aCA
tomoresuitable
CA.
28/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
29/72
ManipulationofCAs
Therearesomeusefuloperationsthat
canbeappliedto
acoveringarrayprevi-
ouslyconstructed.Thissectiondescribes
fourofthem
:maxim
izationofconstant
rows,optimal
reduction,wildcard
detection,andfusion.
�Themaxim
izationofconstantrowsenable
that
theproduct
ofCAs,
and
thepow
eringofCAsproduce
redundantrowsthat
areelim
inated
easily.
�Theoptimal
reductionofCAs,
enable
toconstruct
smallCAstakingas
inputCAswithgreater
number
ofrowsandcolumns.
�Thewildcard
detectionofCAs,will
detectredundantelem
ents
inaCA.
�ThefusionofCAs,enableto
exploitsystem
aticallythewildcardsto
reduce
thesize
ofaCA
30/72
ManipulationofCAs
Maxim
izationofConstantRow
sin
aCA
�A
constantrow
inacoveringarrayisarow
havingthesamesymbolin
allits
elem
ents.Form
ally,thei-th
row
ofacoveringarrayA=
(ai,j)
ofdim
ensions
N×
kisconstantifai,j=
ai,0forj=
1,2
,...,k
−1
13.
�Bymeansofthethreeoperationsthatproduce
isomorphic
coveringarrays
itis
possible
toarrangethesymbolsofthecoveringarrayin
order
tomake
constant
someofitsrows.
�Theconstantrowsarevery
usefulforthemethodsofmultiplicationand
poweringofcoveringarrays,because
ifthecoveringarrays
usedhave
constant
rows,then
itispossible
todeletesomerowsin
theresultingcoveringarray.
�Thisproblem
canbesolved
inthedomain
ofgraphs.
Foreach
row
anodeis
created,andforeach
pairofrowsthataredistinct
(columnbycolumn)anedge
iscreated.Theproblem
isconverted
totheMAXCLIQ
UEproblem.
13Quiz-Ram
os[49]
31/72
0→
02
21
1→
21
10
2→
10
02
3→
10
20
4→
11
01
5→
12
12
6→
20
01
7→
21
22
8→
22
11
9→
00
12
10→
01
10
11→
02
00
38
10
652
01
11
79
4
00
00
11
11
22
22
22
01
21
20
20
12
12
20
11
02
10
10
02
12
01
11
00
21
32/72
ManipulationofCAs
ShorteningofaCA
�Given
acoveringarrayA
theOptimal
ShorteningofCoveringARrays
(OSCAR)problem
consistsin
findingasubmatrixB
ofadetermined
size
such
that
thenumber
ofmissingtuplesin
Bisminim
ized
14.
�Let
beδandΔ
twointegerssuch
that
0≤
δ≤
N−
vt,0≤
Δ≤
k−
t,withtheconditionthat
atleastoneofthem
isgreater
than
zero.
�TheOSCARproblem
consistsin
findingasubmatrixB
ofA
ofsize
(N−δ)
×(k
−Δ)such
that
thenumber
ofmissingtuplesin
Bisminim
al.
�Itwas
provedthat
theOSCARproblem
isNP-Complete
byreducingthe
MAXCOVERproblem
totheOSCARproblem.
�Thesearch
spaceoftheOSCARproblem
is�
NN−δ
��k
k−Δ
� .
14Carrizales-Turrubiates[50]
33/72
00
00
00
00
00
11
11
01
11
00
10
11
01
10
10
10
01
11
10
11
00
CA(6;2,7,2)
00
00
00
00
00
11
11
01
11
00
10
11
01
10
10
10
01
11
10
11
00
Row
andcolumnsto
delete
00
00
00
11
01
01
10
01
11
10
CA(5;2,4,2)
Figure
4:Thedeletionof
onerowandthreecolumnsof
thecovering
arrayCA(6;2,7,2)producesthecoveringarrayCA(5;2,4,2).
34/72
ManipulationofCAs
RedundantElements
Detectionin
aCA
Sometim
esacoveringarrayhasentriesthatcanbefreely
modified
withoutaffecting
thecoverageproperties
ofthecoveringarray,thatis,withoutaffectingthenumber
of
missingtuplesofthearray.
Theseentriesarecalled
wildcardsandarecommonly
representedbythesymbol∗.Figure
5showsattheleft
thecoveringarray
CA(7;2
,8,2),
andshowsatrightthesamecoveringarraywiththewildcardsit
contains.
00
10
10
00
01
00
10
11
11
11
00
10
00
10
11
11
11
01
11
01
10
00
01
00
00
11
00
01
0∗
1∗
10
00
01
00
∗0
1∗
11
11
00
10
00
10
11
11
11
01
11
01
10
00
01
00
00
∗1
00
∗1
Figure
5:Wildcardsin
thecoveringarrayCA(7;2,8,2).
35/72
Wildcard
detectionisvery
importantforsomepostoptimizationprocess
forcovering
arrays.A
postoptimizationprocess
isaprocess
thattriesto
reduce
thenumber
of
rowsofagiven
coveringarray.
Oneoftheseprocess
thatuseswildcardsisthemethod
ofNayeriet
al.
15
Amethodologyto
maximizethenumber
ofwildcardsin
acoveringarray
16proposed
threemain
steps:
�to
determinethetuplescoveredonly
once;
�to
determinetheunfixedsymbols;
�to
enumerate
allthepossible
wildcard
configurations.
Thefirstandsecondstepsareseen
inthenextfigure.Theelem
ents
thatpaticipate
in
t-wisecombinatioscoveredonce
areseen
inthesecondmatrix
as0or1,theunfixed
elem
ents
areshownin
thesecondmatrix
asU.Thethirdstep
canbeim
plemented
usingagreedyapproach
oranexact
approach,butin
anycase
both
approaches
tries
tominim
izetheU
elem
ents
thatbecomefixed(tosomevalueofZv),
andin
thisway
theremainingU
elem
ents
aremaximized
andconverted
towildcards(∗).
15Nayeriet
al.[51]
16Gonzalez-Hernandez
etal.[52]
36/72
00
10
10
00
01
00
10
11
11
11
00
10
00
10
11
11
11
01
11
01
10
00
01
00
00
11
00
01
0U
UU
1U
U0
01
00
U0
1U
11
11
00
10
00
1U
U1
1U
11
01
11
01
10
00
01
U0
00
U1
0U
U1
Figure
6:Theentriesof
thecoveringarrayCA(7;2,8,2)markedwithU
aretheentriesthat
may
becom
eawildcard.
37/72
ManipulationofCAs
FusionofaCA
Colbourn
17defined
thefusionofaCA
as:
CAN(t,k
,v)≤
CAN(t,k
,v+1)−
3ift=
2,k≤
v+
1,visaprimepower
2otherwise
(4)
Thebasicmechanism
ofthefusionoperatoristo
obtain
from
acoveringarrayA=
CA(N
;t,k
,v)another
coveringarrayB
=CA(M
;t,k
,v−
1)ofsm
aller
size
by
replacingtheoccurrencesofthesymbolvin
Aforsymbolsoftheset{0,1
,...,v
−2},
andbydeletingthreeortworowsaccordingthecasesofexpression(4).
17Colbourn
[53],andColbourn
etal.[54]
38/72
Ageneralizationofthefusionoperator
toCAs
18was
proposed.Thegeneraliza-
tionconsistsin
threemaincomponents:
�Awildcard
detector,that
maxim
izes
thewildcardsin
aCA.Theim
plemen-
tationofthiscomponentcould
begreedyor
exact.
�A
row
replacerthat
exploitstheredundantelem
ents
toreduce
therows
oftheCA.Thiscomponentcopiesnon-w
ildcard
elem
ents
tocorresponding
wildcard
elem
ents
(inthesamecolumn)in
order
toobtain
aredundant
row
that
canbeelim
inated.Thisprocedure
could
endwhen
aredundant
row
isfound,or
when
thebestnon-redundantrow
isfound(onerow
that
consumes
theless
number
ofwildcard
elem
ents).
Incase
aredundant
row
could
notbefound,aquasi-CA
with
theminim
um
missingt-wise
combinationsisdelivered.
�A
metaheuristic
algorithm
that
triesto
makezero
thenumber
ofmissing
tuples.
Thisprocedure
isrununtilthenumber
ofmissingtuplesiszero
oramaxim
um
timeisconsumed.
When
theorder
oftheinputCA
isgreater
than
theorder
oftheoutputCA
an
order
reductor
algorithm
isused.
18Rodriguez-Cristerna[55]
39/72
start
order
reductor
wildcard
detector
row
replacer
missing>
0?
simulatedannealing
missing>
0?
end
end
yes
no
yes
yes
no
40/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
41/72
PostprocessigNISTRepository
�AspartoftheAutomated
CombinatorialTestingforSoftware(A
CTS)
project
atNIST,arepository
ofcoveringarrays
ispublicly
available.An
importantopportunityistheexploitationoftheredundancy
ofcoverage
that
thecoveringarrays
havein
order
toreduce
itssize.
�TheapplicationoftheGeneralized
fusionoperator
producesalotof
improvements
oftheCA
oftheNIST
Repository.
�TheNIST
repository
was
processed
usingthegeneralized
fusionandWe
havefound349new
upper
bounds.
�Wehaveem
ployedmorethan
1.5
millionofhours
oftheXiuhcoatlcluster
oftheCINVESTAV.
�A
totalof21,964CAswereprocessed
and20,956wereim
proved.
�Anaveragereductionof3.85%
was
attained.
42/72
0 0.5
1 1.5
2 2.5
3 3.5 2
2.5
3
3.5
4
4.5
5
5.5
6
0
5
10
15
20
25
−Δ
t=2
v2
v3
v4
v5
v6
log(k)
v
−Δ
43/72
0.5
1
1.5
2
2.5
3
3.5 2
2.5
3
3.5
4
4.5
5
5.5
6
0
5
10
15
20
25
30
−Δ
t=3
v2
v3
v4
v5
v6
log(k)
v
−Δ
44/72
0.5
1
1.5
2
2.5
3 2
2.5
3
3.5
4
4.5
5
5.5
6
0
5
10
15
20
25
30
−Δ
t=4
v2
v3
v4
v5
v6
log(k)
v
−Δ
45/72
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2
2.5
3
3.5
4
4.5
5
5.5
6
0
5
10
15
20
25
−Δ
t=5
v2
v3
v4
v5
v6
log(k)
v
−Δ
46/72
0.8
1
1.2
1.4
1.6
1.8
2 2
2.5
3
3.5
4
4.5
5
0
2
4
6
8
10
12
14
16
18
20
−Δ
t=6
v2
v3
v4
v5
log(k)
v
−Δ
47/72
v3t4
v3t4
v6t4
Id.
kIPOG-F
FG-F1
Id.
kIPOG-F
FG-F1
Id.
kIPOG-F
FG-F1
1315
968
964
44
377
1011
1003
85
85
11441
11384
2316
969
964
45
379
1013
1007
86
86
11484
11407
3317
969
963
46
386
1017
1009
87
87
11533
11478
4318
971
965
47
405
1027
1024
88
88
11577
11504
5319
971
963
48
439
1046
1045
89
89
11625
11581
6320
971
963
49
447
1052
1050
90
90
11666
11591
7321
974
965
v6t4
91
91
11710
11630
8322
974
966
Id.
kIPOG-F
FG-F1
92
92
11753
11671
9323
975
970
50
49
9323
9212
93
93
11790
11729
10
324
976
965
51
50
9393
9294
94
94
11833
11762
11
325
976
970
52
51
9466
9397
95
95
11874
11800
12
326
976
974
53
52
9550
9463
96
96
11913
11884
13
327
977
973
54
53
9623
9540
97
97
11956
11918
14
328
978
969
55
55
9762
9673
98
98
11997
11924
15
329
979
975
56
56
9828
9742
99
99
12038
11949
16
330
979
976
57
57
9900
9813
100
100
12085
12003
17
331
981
977
58
58
9964
9869
101
101
12120
12036
18
332
981
973
59
59
10032
9948
102
102
12148
12140
19
333
983
975
60
60
10097
10013
103
103
12194
12120
20
335
983
979
61
61
10163
10067
104
104
12231
12185
21
336
983
978
62
62
10219
10142
105
105
12267
12196
22
337
984
977
63
63
10282
10198
106
106
12306
12240
23
338
985
977
64
64
10347
10250
107
107
12343
12268
24
339
986
977
65
65
10398
10328
108
109
12408
12388
25
340
987
978
66
66
10463
10368
109
110
12445
12380
26
341
987
979
67
67
10520
10441
110
112
12517
12449
27
342
987
981
68
68
10578
10478
111
116
12651
12593
28
343
990
985
69
69
10633
10572
112
118
12716
12698
29
345
991
984
70
70
10693
10599
113
120
12785
12734
30
346
992
985
71
71
10745
10676
114
121
12816
12757
31
347
992
985
72
72
10798
10744
115
128
13036
13007
32
348
992
987
73
73
10850
10758
116
129
13062
13056
33
349
992
986
74
74
10909
10821
117
133
13192
13146
34
351
993
991
75
75
10958
10882
118
149
13634
13606
35
353
994
987
76
76
11012
10959
v3t5
36
360
999
994
77
77
11057
10992
Id.
kIPOG-F
FG-F1
37
361
1001
995
78
78
11110
11017
119
35
1867
1826
38
362
1001
996
79
79
11158
11100
120
36
1895
1850
39
364
1002
997
80
80
11203
11121
121
37
1920
1882
40
368
1007
1002
81
81
11253
11187
122
38
1947
1909
41
370
1008
1001
82
82
11303
11219
123
39
1974
1933
42
373
1009
1003
83
83
11353
11282
124
40
1997
1949
43
375
1009
1005
84
84
11397
11319
125
41
2023
1975
47/72
v3t5
v3t5
v4t5
Id.
kIPOG-F
FG-F1
Id.
kIPOG-F
FG-F1
Id.
kIPOG-F
FG-F1
126
42
2046
2002
169
85
2739
2700
210
45
9227
9104
127
43
2070
2022
170
86
2749
2706
211
46
9320
9176
128
44
2091
2050
171
87
2762
2728
212
47
9406
9262
129
45
2112
2071
172
88
2770
2736
213
48
9501
9357
130
46
2130
2086
173
89
2783
2747
214
49
9588
9453
131
47
2150
2112
174
90
2792
2761
215
50
9673
9520
132
48
2174
2134
175
91
2805
2772
216
51
9755
9621
133
49
2191
2154
176
92
2815
2783
217
52
9835
9682
134
50
2213
2182
177
93
2825
2798
218
53
9922
9769
135
51
2231
2198
178
94
2836
2797
219
54
9998
9849
136
52
2251
2217
179
95
2847
2813
220
55
10079
9927
137
53
2269
2232
180
96
2857
2819
221
56
10155
10007
138
54
2289
2246
181
97
2868
2834
222
57
10232
10082
139
55
2309
2277
182
98
2877
2838
223
59
10379
10231
140
56
2327
2281
183
99
2885
2854
224
60
10454
10302
141
57
2342
2304
184
100
2895
2862
225
62
10590
10442
142
58
2358
2316
185
101
2909
2871
226
63
10650
10509
143
59
2373
2335
186
102
2920
2878
v5t5
144
60
2394
2347
187
103
2928
2898
Id.
kIPOG-F
FG-F1
145
61
2408
2370
188
104
2938
2901
227
21
18779
18260
146
62
2425
2387
189
105
2945
2908
228
22
114775
111818
147
63
2440
2397
190
107
2962
2928
229
23
118587
115802
148
64
2451
2413
v4t5
230
24
122201
119500
149
65
2468
2439
Id.
kIPOG-F
FG-F1
231
25
125683
123108
150
66
2482
2447
191
26
6957
6775
v6t5
151
67
2498
2459
192
27
7116
6937
Id.
kIPOG-F
FG-F1
152
68
2512
2477
193
28
7267
7088
232
17
40334
38976
153
69
2527
2487
194
29
7414
7247
233
18
42102
40820
154
70
2542
2504
195
30
7555
7379
234
19
43833
42554
155
71
2555
2518
196
31
7691
7527
235
20
45425
44224
156
72
2573
2531
197
32
7816
7649
236
21
46970
45784
157
73
2584
2546
198
33
7939
7782
237
22
48479
47352
158
74
2597
2564
199
34
8064
7907
238
23
49924
48838
159
75
2609
2588
200
35
8183
8023
239
24
51287
50180
160
76
2625
2593
201
36
8301
8137
240
25
52604
51505
161
77
2639
2608
202
37
8420
8259
241
26
53850
52814
162
78
2648
2615
203
38
8530
8376
242
27
55069
54032
163
79
2661
2625
204
39
8629
8477
243
28
56225
55275
164
80
2673
2635
205
40
8737
8583
244
29
57363
56353
165
81
2686
2651
206
41
8847
8693
245
30
58468
57503
166
82
2700
2668
207
42
8945
8791
246
31
59529
58576
167
83
2710
2677
208
43
9035
8890
247
32
60570
59612
168
84
2725
2684
209
44
9137
8979
248
33
61562
60608
47/72
v6t5
v2t6
v4t6
Id.
kIPOG-F
FG-F1
Id.
kIPOG-F
FG-F1
Id.
kIPOG-F
FG-F1
249
34
62527
61612
290
79
782
770
329
26
33369
32513
250
35
63471
62557
291
80
785
771
330
27
34187
33356
251
36
64399
63519
292
81
791
772
331
28
35006
34198
v2t6
293
82
794
778
332
29
35791
34971
Id.
kIPOG-F
FG-F1
294
83
796
787
333
30
36570
35778
252
41
572
556
295
84
800
784
334
31
37305
36515
253
42
579
561
296
85
804
790
335
32
38015
37255
254
43
590
570
297
86
809
799
v5t6
255
44
594
580
v3t6
Id.
kIPOG-F
FG-F1
256
45
603
589
Id.
kIPOG-F
FG-F1
336
11
56615
52471
257
46
611
596
298
26
5709
5544
337
12
63620
59622
258
47
617
602
299
27
5853
5667
338
13
70190
66275
259
48
625
612
300
28
6003
5827
339
14
76390
72680
260
49
630
617
301
29
6150
5969
340
15
82139
78480
261
50
636
623
302
30
6281
6103
341
16
87559
84102
262
51
643
629
303
31
6413
6245
342
18
97605
94263
263
52
650
636
304
32
6535
6348
343
19
102208
98994
264
53
656
641
305
33
6656
6461
344
20
106642
103514
265
54
662
644
306
34
6772
6583
345
21
110842
107773
266
55
667
653
307
35
6877
6715
346
22
114775
111818
267
56
672
655
308
36
6989
6832
347
23
118587
115802
268
57
677
663
309
37
7092
6932
348
24
122201
119500
269
58
683
667
310
38
7194
7036
349
25
125683
123108
270
59
689
671
311
39
7293
7131
271
60
695
680
312
40
7391
7233
272
61
699
687
313
41
7490
7315
273
62
703
686
314
42
7574
7411
274
63
709
697
315
43
7672
7506
275
64
715
699
316
44
7757
7600
276
65
721
704
317
45
7845
7702
277
66
725
709
318
46
7938
7766
278
67
728
710
319
47
8013
7856
279
68
732
718
320
50
8256
8108
280
69
738
724
321
51
8333
8179
281
70
743
729
v4t6
282
71
747
734
Id.
kIPOG-F
FG-F1
283
72
751
736
322
19
26392
25430
284
73
755
743
323
20
27534
26564
285
74
761
749
324
21
28625
27676
286
75
766
751
325
22
29640
28735
287
76
770
755
326
23
30636
29720
288
77
773
758
327
24
31591
30724
289
78
777
760
328
25
32501
31654
47/72
TableofContents
Introduction
BasicDefinitions
CARepositories
ConstructionMethods
Manipulation
ofCAs
PostprocessingNISTRepository
Conclusions
48/72
Conlusions
�TheCoveringArrays(C
A)representanexcelentoptionwhen
itisdesired
toguaranteeacertain
levelofcoveragein
thetestingofasoftwarecomponentand
touse
asm
allnumber
oftest
cases.
�Wehave
presentedbasicdefinitionsneeded
tounderstandmuch
oftheresearch
relatedto
coveringarrays.
�Wehave
grouped
alotofresearch
reportsrelatedto
theconstructionofCAs,
thegroupswere:
Exact
methods,GreedyMethods,Metaheuristic
Methods,and
ManipulationMethods.
�Thegeneralizedfusionoperatorisbasedontheexploitationofthewildcardsof
aCA
inorder
toreduce
thenumber
ofrowsoftheCA.
�Wehave
processed
alltheNIST
CA
Repository
usingmore
than1.5
millonof
CPU
hours
(XiuhcoatlCinvestavCluster)andhave
found349new
upper
bounds.
�Itisim
portantto
highlightthattheexploitationoftheredundancy
ofcoverage
inCAscanenable
thecreationofbetterCAs(w
ithlower
number
ofrows).
49/72
ReferencesI
[1]
B.HailpernandP.Santhanam
,“Softwaredebugging,
testing,
andverification,”
IBM
SystemsJournal,vol.41,no.
1,pp.4
–12,
2002.
[2]
A.Hartm
an,“Softwareandhardwaretestingusing
combinatorialcoveringsuites,”
inGraphTheory,
CombinatoricsandAlgorithms,ser.Operations
Research/C
omputerScience
Interfaces
Series.
SpringerUS,
2005,vol.34,pp.237–266.
[3]
D.Cohen,S.Dalal,J.
Parelius,andG.Patton,“T
he
combinatorialdesignapproach
toautomatictest
generation,”
Software,
IEEE,vol.13,no.
5,pp.83
–88,
sep1996.
[4]
D.Cohen,S.Dalal,M.Fredman,andG.Patton,“T
heaetg
system
:Anapproach
totestingbased
oncombinatorial
design,”
IEEETransactionson
SoftwareEngineering,
vol.23,
no.
7,pp.437–444,
jul1997.
51/72
ReferencesII
[5]
K.Bush,“O
rthogonalarrays
ofindex
unity,”TheAnnalsof
Mathem
atical
Statistics,vol.23,no.
3,pp.426–434,1952.
[6]
J.Torres-Jimenez,N.Rangel-Valdez,L.Gonzalez-Hernandez,
andH.Avila-George,“C
onstructionof
logarithm
tablesfor
GaloisFields,”International
Journal
ofMathem
atical
Education
inScience
andTechnology,vol.42,no.
1,pp.
91–102,Oct.2010.
[7]
CryptographyandSecurity
inCom
puting.
InTech,2012,ch.
Constructionof
OrthogonalArraysof
Index
UnityUsing
Logarithm
TablesforGaloisFields.
[8]
C.Colbourn,“C
ombinatorialaspects
ofcoveringarrays,”
Le
Matem
atiche,vol.59,no.
1,2,
pp.125–172,
2004.
[9]
Foundationsof
Probability.
Holden-Day,1970.
52/72
ReferencesIII
[10]G.Katona,
“Twoapplications(for
search
theory
andtruth
functions)
ofsperner
typetheorems,”PeriodicaMathem
atica
Hungarica,1973.
[11]
D.Kleitman
andJ.
Spencer,“Fam
ilies
ofk-independent
sets,”
DiscreteMathem
atics,1973.
[12]
D.Kuhn,J.
Higdon,J.
Law
rence,R.Kacker,andL.Y.,
“Com
binatorialmethodsforeventsequence
testing,”in
Workshop
onCom
binatorialTesting(CT)In
conjunctionwith
International
Conference
onSoftwareTesting.
[13]J.
Yan
andJ.
Zhang,
“Backtrackingalgorithmsandsearch
heuristicsto
generatetest
suites
forcombinatorialtesting,”in
Proceedings
ofthe30th
Annual
International
Com
puter
SoftwareandApplicationsConference
-Volume01,ser.
COMPSAC’06.
Washington,DC,USA:IEEECom
puter
53/72
ReferencesIV
Society,2006,pp.385–394.
[Online].Available:
http://dx.doi.org/10.1109/C
OMPSAC.2006.33
[14]J.
Bracho-Rios,J.Torres-Jimenez,andE.Rodriguez-Tello,“A
new
backtrackingalgorithm
forconstructingbinarycovering
arrays
ofvariablestrength,”
inMICAI2009:Advancesin
Artificial
Intelligence,ser.Lecture
Notes
inCom
puterScience,
A.Aguirre,R.Borja,andC.Garcia,
Eds.
SpringerBerlin
/Heidelberg,
2009,vol.5845,pp.397–407.
[Online].Available:
http://dx.doi.org/10.1007/978-3-642-05258-3
35
[15]B.Hnich,S.D.Prestwich,E.Selensky,andB.M.Smith,
“Constraintmodelsforthecoveringtest
problem,”
Constraints,vol.11,no.
2-3,
pp.199–219,
Jul.2006.[Online].
Available:http://dx.doi.org/10.1007/s10601-006-7094-9
54/72
ReferencesV
[16]D.Lopez-Escogido,
J.Torres-Jimenez,E.Rodriguez-Tello,
andN.Rangel-Valdez,“Strengthtwocoveringarrays
constructionusingasatrepresentation,”
inMICAI,2008,pp.
44–53.
[17]M.Banbara,
H.Matsunaka,
N.Tam
ura,andK.Inoue,
“Generatingcombinatorialtest
casesby
efficientsat
encodings
suitableforcdclsatsolvers,”in
Proceedings
ofthe
17th
international
conference
onLogicforprogramming,
artificial
intelligence,andreasoning,
ser.LPAR’10.
Berlin,
Heidelberg:
Springer-Verlag,
2010,
pp.112–126.
[18]K.Meagher,“N
on-isomorphicgenerationof
coveringarrays,”
University
ofRegina,
Tech.Rep.,2002.
[19]R.Bryce
andC.Colbourn,“T
hedensity
algorithm
for
pairwiseinteractiontesting,”SoftwareTesting,
Verification
andReliability,vol.17,no.
3,pp.159–182,
2007.
55/72
ReferencesVI
[20]Y.Lei,R.Kacker,D.Kuhn,V.Okun,andJ.Law
rence,“Ipog:
ageneralstrategy
fort-way
softwaretesting,”in
Proceedings
ofthe14th
Annual
IEEEInternational
Conference
and
Workshopson
theE
ngineeringof
Com
puter-Based
Systems.
ECBS’07.,Tucson,AZ,USA,mar
2007,pp.549–556.
[21]
M.Forbes,J.
Law
rence,Y.Lei,R.Kacker,andD.Kuhn,
“Refiningthein-param
eter-order
strategy
forconstructing
coveringarrays,”
Journal
ofResearchof
theNational
Institute
ofStandardsandTechnology,vol.113,
no.
5,pp.287–297,
2008.
56/72
ReferencesVII
[22]Y.Lei,R.Kacker,D.R.Kuhn,V.Okun,andJ.
Law
rence,
“Ipog:Ageneralstrategy
fort-way
softwaretesting,”in
Proceedings
ofthe14th
Annual
IEEEInternational
Conference
andWorkshopson
theEngineeringof
Com
puter-Based
Systems,ser.ECBS’07.
Washington,DC,
USA:IEEECom
puterSociety,2007,pp.549–556.
[Online].
Available:http://dx.doi.org/10.1109/E
CBS.2007.47
[23]
A.CalvagnaandA.Gargantini,“T
-wisecombinatorial
interactiontest
suites
constructionbased
oncoverage
inheritance,”
SoftwareTesting,
Verification
andReliability,
vol.22,no.
7,pp.507–526,
2012.[Online].Available:
http://dx.doi.org/10.1002/stvr.466
57/72
ReferencesVIII
[24]E.Covarrubias-Flores,“C
alculodecoveringarrays
binariosde
fuerza
variable,usandounalgoritm
oderecocidosimulado,”
Master’sthesis,CentrodeInvestigacionydeEstudios
Avanzados
delInstituto
PolitecnicoNacional,2008.
[25]J.
Torres-Jimenez
andE.Rodriguez-Tello,“N
ewboundsfor
binarycoveringarrays
usingsimulatedannealing,”Inform
ation
Sciences,vol.185,
no.
1,pp.137–152,
February2012.
[26]
H.Avila-George,J.
Torres-Jimenez,andV.Hernandez,
“Parallelsimulatedannealingforthecoveringarrays
constructionproblem,”
inToappearin
the18th
International
Conference
onParallelandDistributedProcessingTechniques
andApplications,2012.
58/72
ReferencesIX
[27]H.Avila-George,J.
Torres-Jimenez,andV.Hernandez,“N
ewboundsforternarycoveringarrays
usingaparallelsimulated
annealing,”Mathem
atical
Problemsin
Engineering,
Accepted
onJuly07,2012.
[28]K.Nurm
ela,
“Upper
boundsforcoveringarrays
bytabu
search,”
DiscreteAppliedMathem
atics,vol.138,
no.
1-2,
pp.
143–152,
2004,optimalDiscreteStructuresandAlgorithms.
[29]
L.Gonzalez-Hernandez,N.Rangel-Valdez,and
J.Torres-Jimenez,“C
onstructionof
mixed
coveringarrays
ofvariablestrengthusingatabusearch
approach,”
inCombinatorialOptimizationandApplications,ser.Lecture
Notes
inCom
puterScience,W.WuandO.Daescu,Eds.
SpringerBerlin
/Heidelberg,
2010,vol.6508,pp.51–64.
[Online].Available:
http://dx.doi.org/10.1007/978-3-642-17458-2
6
59/72
ReferencesX
[30]J.
Stardom
,“M
etaheuristicsandthesearch
forcoveringand
packingarrays,”
Master’sthesis,Simon
FraserUniversity,
2001.
[31]T.Shiba,
T.Tsuchiya,
andT.Kikuno,
“Usingartificiallife
techniques
togeneratetest
casesforcombinatorialtesting,”
inCom
puterSoftwareandApplicationsConference,2004.
COMPSAC2004.Proceedings
ofthe28th
Annual
International,vol.1,
sept.2004,
pp.72–77.
[32]
D.T.TangandL.S.Woo,
“Exhaustivetest
pattern
generationwithconstantweightvectors,”IEEETrans.
Comput.,vol.32,no.
12,pp.1145–1150,
Dec.1983.[Online].
Available:http://dx.doi.org/10.1109/T
C.1983.1676175
60/72
ReferencesXI
[33]J.
Martinez-PenaandJ.
Torres-Jimenez,“A
branch
and
boundalgorithm
forternarycoveringarrays
constructionusing
trinom
ialcoeffi
cients,”
Researchin
Com
putingScience,
vol.49,pp.61–71,
2010.
[34]J.
Martinez-Pena,
J.Torres-Jimenez,N.Rangel-Valdez,and
H.Avila-George,“A
heuristicapproach
forconstructing
ternarycoveringarrays
usingtrinom
ialcoeffi
cients,”
inProceedings
ofthe12th
Ibero-American
conference
onAdvancesin
artificial
intelligence,ser.IBERAMIA’10.
Berlin,
Heidelberg:
Springer-Verlag,
2010,
pp.572–581.
[Online].
Available:
http://dl.acm.org/citation.cfm
?id=1948131.1948203
[35]C.Colbourn,“C
overingarrays
from
cyclotom
y,”Designs,
Codes
andCryptography,vol.55,pp.201–219,
2010.[Online].
Available:http://dx.doi.org/10.1007/s10623-009-9333-8
61/72
ReferencesXII
[36]K.Meagher
andB.Stevens,“G
roupconstructionof
covering
arrays,”
Journal
ofCom
binatorialDesigns,vol.13,no.
1,pp.
70–77,2005.[Online].Available:
http://dx.doi.org/10.1002/jcd.20035
[37]
J.Lobb,C.Colbourn,P.Danziger,B.Stevens,and
J.Torres-Jimenez,“C
over
startersforcoveringarrays
ofstrengthtwo,”DiscreteMathem
atics,2011,in
press.
[38]
N.J.
A.Sloane,“C
overingarrays
andintersectingcodes,”
Journal
ofCom
binatorialDesigns,vol.1,
no.
1,pp.51–63,
1993.
[39]
G.Roux,
“k-proprietesdansdes
tableauxdencolonnes;cas
particulierdelak-surjectivite
etdelak-permutivite,”
Ph.D.
dissertation,University
ofParis,1987.
62/72
ReferencesXIII
[40]M.ChateauneufandD.Kreher,“O
nthestateof
strength-threecoveringarrays,”
Journal
ofCombinatorial
Designs,vol.10,no.4,
pp.217–238,2002.
[41]M.B.Cohen,C.J.
Colbourn,andA.C.Lin,“C
onstructing
strengththreecoveringarrays
withaugm
entedannealing,”
DiscreteMathem
atics,vol.308,
no.
13,pp.2709–2722,2008.
[42]
C.J.
Colbourn,S.S.Martirosyan,T.Trung,
andR.A.
Walker,II,“R
oux-typeconstructionsforcoveringarrays
ofstrengthsthreeandfour,”Des.Codes
Cryptography,vol.41,
no.1,
pp.33–57,
Oct.2006.[Online].Available:
http://dx.doi.org/10.1007/s10623-006-0020-8
[43]S.Martirosyan
andT.v.
Trung,
“Ont-coveringarrays,”
Designs,Codes
andCryptography,vol.32,pp.323–339,2004,
10.1023/B:DESI.0000029232.40302.6d.[Online].Available:
http://dx.doi.org/10.1023/B
:DESI.0000029232.40302.6d
63/72
ReferencesXIV
[44]C.J.
Colbourn,S.S.Martirosyan,G.L.Mullen,D.Shasha,
G.B.Sherwood,andJ.
L.Yucas,“P
roductsof
mixed
coveringarrays
ofstrengthtwo,”Journal
ofCom
binatorial
Designs,vol.14,no.2,
pp.124–138,2006.
[45]
A.Hartm
an,“Softwareandhardwaretestingusing
combinatorialcoveringsuites,”
inGraphTheory,
Com
binatoricsandAlgorithms,ser.Operations
Research/C
omputerScience
Interfaces
Series,M.C.Golumbic
andI.B.-A.Hartm
an,Eds.
SpringerUS,2005,vol.34,pp.
237–266.[Online].Available:
http://dx.doi.org/10.1007/0-387-25036-0
10
[46]
H.Avila-George,“V
erificacion
decoveringarrays
utilizando
supercomputacion
ycomputacion
grid,”
Master’sthesis,
Universidad
PolitecnicadeValencia,
2010.
64/72
ReferencesXV
[47]H.Avila-George,J.
Torres-Jimenez,V.Hernandez,and
N.Rangel-Valdez,“A
parallelalgorithm
fortheverification
ofcoveringarrays,”
inProceedings
ofTheInternational
Conference
onParallelandDistributedTechniques
and
Applications(PDPTA2011),Las
Vegas
Nevada,
USA,18-21
July2011,pp.879–885.
[Online].Available:
http://w
orld-com
p.org/p2011/P
DP8061.pdf
[48]——,“V
erification
ofgeneralandcycliccoveringarrays
using
grid
computing,”in
Proceedings
oftheThirdinternational
conference
onDatamanagem
entin
grid
andpeer-to-peer
system
s,ser.Globe’10.
Berlin,Heidelberg:
Springer-Verlag,
2010,pp.112–123.
[Online].Available:
http://dl.acm.org/citation.cfm
?id=1885229.1885242
65/72