Co esponding au ho : Ramchande Malkoochi.
Copy igh © 2025 Au ho (s) e ain he copy igh o his a icle. This a icle is published unde he e ms o he C ea i e Commons A ibu ion License 4.0.
Mul ilaye ed CBDC F aud De ec ion F amewo k: Secu ing Digi al Cu ency
Ecosys ems
Ramchande Malkoochi *
Malaysia uni e si y, Malaysia.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
Publica ion his o y: Recei ed on 03 Ap il 2025; e ised on 11 May 2025; accep ed on 13 May 2025
A icle DOI: h ps://doi.o g/10.30574/wja .2025.26.2.1841
Abs ac
This a icle shows he de elopmen and implemen a ion o aud moni o ing sys ems o Cen al Bank Digi al
Cu encies (CBDCs). The a icle ackles he dis inc i e secu i y issues associa ed wi h digi al cu encies by employing a
comp ehensi e me hodology ha in eg a es a sys ema ic li e a u e e iew, de ailed case s udy examina ion, s a is ical
analysis o ansac ion da ase s, and igo ous alida ion es ing p o ocols. The a icle iden i ies c i ical ulne abili ies
in exis ing inancial sys ems and p oposes a mul i-laye ed de ense amewo k ha balances secu i y impe a i es wi h
p i acy conside a ions. S a is ical analysis demons a es ha he p oposed a chi ec u e achie es supe io aud
de ec ion capabili ies while main aining ope a ional e iciency ac oss di e se a ack ec o s. The a icle explo es
ensions be ween p i acy p ese a ion and comp ehensi e moni o ing, examines scalabili y conce ns o la ge-scale
implemen a ions, analyzes c oss-bo de ansac ion complexi ies, and e alua es egula o y amewo ks o CBDC aud
in es iga ion. Fu u e di ec ions emphasize he in eg a ion o ad anced AI echniques, collabo a i e amewo ks o
c oss-ju isdic ional moni o ing, p i acy-enhancing echnologies, s anda diza ion oppo uni ies, and s a egic esea ch
p io i ies o add ess emaining gaps in CBDC secu i y capabili ies.
Keywo ds: CBDC Secu i y; F aud De ec ion; P i acy-P ese ing Moni o ing; C oss-Bo de T ansac ions; A i icial
In elligence
1. In oduc ion
Cen al Bank Digi al Cu encies (CBDCs) ep esen one o he mos signi ican inno a ions in mone a y sys ems since
he ansi ion om me allic o ia cu ency. As o 2024, app oxima ely 134 coun ies, ep esen ing o e 98% o global
GDP, a e ac i ely explo ing CBDCs, wi h 11 coun ies ha ing ully launched digi al cu encies and 21 in ad anced pilo
phases [1]. Cu en pilo p og ams ha e p ocessed o e 950 million ansac ions wo h app oxima ely 220 billion in
equi alen alue since hei ini ial launches, while majo economic egions ha e engaged ens o housands o use s in
con olled es ing en i onmen s ac oss mul iple ju isdic ions [1].
The digi aliza ion o so e eign cu encies in oduces unique secu i y challenges dis inc om hose in adi ional
banking sys ems. Unlike physical cash, which equi es physical p esence o he , CBDCs exis in digi al en i onmen s
whe e sophis ica ed cybe h ea s can ope a e ac oss ju isdic ions. Recen da a indica es ha inancial ins i u ions
expe ience 300% mo e cybe a acks han o ganiza ions in o he sec o s, wi h a acks speci ically a ge ing digi al
paymen sys ems inc easing by 56% be ween 2021 and 2023 [1]. The po en ial concen a ion o inancial da a wi hin
CBDC sys ems c ea es high- alue a ge s o malicious ac o s, necessi a ing obus secu i y a chi ec u es ha can
wi hs and e ol ing h ea landscapes.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
1996
These eme ging h ea s unde sco e he necessi y o specialized aud moni o ing sys ems o CBDCs. T adi ional aud
de ec ion mechanisms employed in elec onic banking ely p edominan ly on pa e n ecogni ion wi hin limi ed
ansac ion se s and o en ope a e wi h signi ican p ocessing la ency. Howe e , CBDCs equi e eal- ime moni o ing
capabili ies ac oss po en ially billions o daily ansac ions while main aining sub-second esponse imes. Acco ding o
echnical speci ica ions om pilo CBDC p og ams, hese sys ems mus achie e 99.99% up ime eliabili y while
p ocessing up o 300,000 ansac ions pe second du ing peak pe iods—pe o mance me ics ha exceed hose o mos
cu en comme cial paymen ne wo ks by an o de o magni ude [2].
Cu en aud de ec ion in adi ional banking sys ems p ima ily employs ule-based app oaches supplemen ed by
basic machine lea ning algo i hms. These sys ems demons a e de ec ion a es o app oxima ely 70-85% o known
aud pa e ns bu signi ican ly lowe a es (15-40%) o no el a ack ec o s [2]. Addi ionally, alse posi i e a es in
adi ional sys ems ange om 2-5%, esul ing in subs an ial ope a ional cos s and cus ome ic ion. A 2023 su ey o
78 cen al inancial ins i u ions conduc ed by in e na ional mone a y au ho i ies e ealed ha 63% conside exis ing
aud de ec ion amewo ks inadequa e o CBDC implemen a ion, wi h pa icula conce ns ega ding p i acy-
p ese ing moni o ing and c oss-bo de ansac ion o e sigh [2].
This esea ch aims o add ess hese c i ical gaps by de eloping a comp ehensi e CBDC aud moni o ing amewo k
ha balances secu i y impe a i es wi h ope a ional equi emen s and p i acy conside a ions. The s udy's signi icance
lies in i s po en ial o es ablish a chi ec u al s anda ds and echnical speci ica ions o secu e CBDC implemen a ions.
By ad ancing he unde s anding o digi al cu ency secu i y equi emen s, his esea ch con ibu es o he b oade
objec i e o enabling us ed, esilien digi al inancial ecosys ems capable o suppo ing he u u e o money. The
indings a e in ended o in o m bo h echnological de elopmen and policy o ma ion as inancial au ho i ies wo ldwide
ad ance hei CBDC ini ia i es.
2. Resea ch me hodology
This s udy employs a comp ehensi e mul i-me hod app oach o de elop and alida e a aud moni o ing sys em o
Cen al Bank Digi al Cu encies (CBDCs). The esea ch me hodology in eg a es sys ema ic e iew p ocedu es, case
s udy in es iga ions, quan i a i e da a analysis, amewo k de elopmen , and alida ion es ing o ensu e bo h
heo e ical igo and p ac ical applicabili y o he indings. [3]
A sys ema ic e iew o exis ing CBDC implemen a ions and pilo s was conduc ed ollowing he PRISMA (P e e ed
Repo ing I ems o Sys ema ic Re iews and Me a-Analyses) p o ocol. The e iew encompassed 24 CBDC p ojec s
ac oss 18 ju isdic ions, ep esen ing a ious a chi ec u al app oaches and implemen a ion s ages. Da a ex ac ion
ocused on secu i y amewo ks, aud de ec ion mechanisms, and epo ed secu i y inciden s. The analysis e ealed
ha 73% o ad anced CBDC pilo s u ilize a wo- ie dis ibu ion model, while 68% implemen some o m o ze o-
knowledge p oo echnology o p i acy-p ese ing ansac ion alida ion. Addi ionally, 47% o p ojec s explici ly
inco po a e AI-based anomaly de ec ion, hough implemen a ion de ails a y signi ican ly. The sys ema ic e iew
iden i ied a c i ical esea ch gap: while 91% o CBDC p ojec s men ion aud p e en ion as a p io i y, only 23% p o ide
speci ic echnical documen a ion on hei aud de ec ion sys ems. [3]
The esea ch me hodology inco po a ed a de ailed case s udy analysis o aud pa e ns obse ed in exis ing digi al
paymen sys ems as p oxies o po en ial CBDC ulne abili ies. Six majo digi al paymen ecosys ems we e examined,
documen ing 1,248 unique aud inciden s occu ing be ween 2019-2023. These cases we e ca ego ized using a
s uc u ed axonomy, e ealing ha accoun akeo e a acks (36.4%), social enginee ing schemes (27.8%), and
echnical exploi ec o s (22.1%) cons i u ed he p ima y aud ca ego ies. Tempo al analysis indica ed a 143%
inc ease in API-based a acks a ge ing paymen in as uc u e be ween 2020 and 2023, while c eden ial he a emp s
h ough phishing inc eased by 89% du ing he same pe iod. No ably, c oss-bo de ansac ions exhibi ed ulne abili y
a es 2.6 imes highe han domes ic ansac ions, highligh ing a c i ical a ea o CBDC secu i y enhancemen . [3]
Quan i a i e analysis o ansac ion da a om CBDC es en i onmen s o med he empi ical ounda ion o his esea ch.
Th ough da a-sha ing ag eemen s wi h h ee cen al banks conduc ing CBDC pilo s, he s udy analyzed 17.8 million
anonymized ansac ions om con olled es en i onmen s. These da ase s we e supplemen ed wi h syn he ic
ansac ion da a gene a ed using agen -based simula ion models calib a ed o ma ch obse ed beha io al pa e ns. The
combined da ase was analyzed using mul i a ia e s a is ical echniques o iden i y ansac ion anomalies and es ablish
baseline me ics o no mal beha io ac oss di e en use segmen s and ansac ion ypes. The analysis e ealed ha
ansac ion eloci y, ne wo k cen ali y measu es, empo al pa e ns, and geog aphic dis ibu ion se ed as he mos
signi ican indica o s o aud de ec ion, collec i ely accoun ing o 82.3% o he disc imina o y powe in iden i ying
suspicious ac i i ies. [4]
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
1997
Building on hese analy ical ounda ions, he s udy de eloped a comp ehensi e aud isk assessmen amewo k o
CBDCs h ough an i e a i e design science app oach. The amewo k consis s o i e in e connec ed componen s: (1) a
mul i-laye ed de ec ion a chi ec u e combining ule-based and machine-lea ning app oaches; (2) a isk-sco ing
mechanism inco po a ing 37 weigh ed a iables; (3) a decision ma ix o escala ion p ocedu es; (4) an adap i e
lea ning sys em o con inuous imp o emen ; and (5) p i acy-p ese ing moni o ing p o ocols. The amewo k
de elopmen p ocess included h ee design i e a ions, each e ined h ough expe eedback om 18 specialis s in
cybe secu i y, paymen sys ems, and cen al banking h ough a modi ied Delphi app oach. The inal amewo k
achie ed consensus app o al om 89% o he expe panel, wi h pa icula s eng h acknowledged in i s abili y o
balance secu i y equi emen s wi h p i acy conside a ions. [4]
The me hodology concluded wi h igo ous alida ion using simula ed a ack scena ios. A ed eam o secu i y
p o essionals conduc ed 42 dis inc a ack simula ions agains he de eloped moni o ing sys em, encompassing bo h
known aud ypologies and no el a ack ec o s. These simula ions we e pe o med in a sandboxed en i onmen
eplica ing he echnical a chi ec u e o a wo- ie CBDC sys em. Pe o mance me ics we e e alua ed based on
de ec ion a es, alse posi i e a ios, ime- o-de ec ion, and sys em esilience. The alida ion es ing demons a ed ha
he p oposed moni o ing sys em success ully de ec ed 94.3% o simula ed a acks, wi h an a e age de ec ion ime o
2.7 seconds and a alse posi i e a e o 0.8%. Sys em pe o mance emained s able unde s ess condi ions, simula ing
ansac ion olumes o up o 500,000 ope a ions pe second, con i ming he scalabili y o he app oach o ull-scale
CBDC deploymen s. [4]
Figu e 1 Bibliog aphy P ocedu e o CBDC F aud Moni o ing Sys em S udy [3, 4]
3. S a is ics
The s a is ical analysis o aud moni o ing sys ems o Cen al Bank Digi al Cu encies (CBDCs) p o ides c i ical
insigh s in o de ec ion e ec i eness, pe o mance benchma ks, and compa a i e ad an ages o e adi ional paymen
sys ems. This sec ion p esen s comp ehensi e s a is ical e idence de i ed om bo h empi ical s udies o exis ing digi al
paymen ecosys ems and con olled expe imen al e alua ions o p o o ype CBDC aud moni o ing implemen a ions.
[5]
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
1998
Compa a i e analysis o aud a es be ween adi ional and digi al cu encies e eals signi ican a ia ions in
ulne abili y p o iles and a ack ec o s. Based on consolida ed da a om 15 ju isdic ions spanning 2020-2024,
adi ional ca d-based paymen sys ems expe ience an a e age aud a e o 7.32 basis poin s (0.0732%) by ansac ion
alue, while digi al walle sys ems demons a e a sligh ly lowe a e a 6.98 basis poin s (0.0698%). In con as ,
p elimina y da a om con olled CBDC pilo s indica es subs an ially lowe aud a es, a e aging 1.45 basis poin s
(0.0145%) ac oss implemen a ion ypes. This 79.3% educ ion in aud a es can be a ibu ed o he a chi ec u al
ad an ages o CBDCs, including c yp og aphic ansac ion alida ion, ampe -e iden dis ibu ed ledge s, and enhanced
au hen ica ion mechanisms. Howe e , analysis o a ack ec o dis ibu ions shows conce ning ends: while adi ional
paymen sys ems p ima ily ace ca d-no -p esen aud (59.3% o inciden s) and coun e ei a emp s (24.7%), CBDC
es en i onmen s ha e encoun e ed mo e sophis ica ed h ea s, wi h c eden ial-based a acks (38.9%) and consensus
mechanism exploi s (31.6%) eme ging as p ima y ulne abili y ca ego ies. The s a is ical signi icance o hese
dis ibu ion di e ences is con i med h ough chi-squa e analysis (p < 0.001), indica ing dis inc secu i y challenges o
CBDC implemen a ions despi e hei o e all aud a e ad an ages. [5]
T ansac ion anomaly de ec ion ep esen s a co ne s one o e ec i e CBDC aud moni o ing, wi h s a is ical me ics
e ealing signi ican pe o mance a ia ions ac oss me hodologies. E alua ion o 23 de ec ion algo i hms agains a
common da ase o 4.7 million ansac ions (including 2,340 known audulen examples) demons a es ha ensemble
app oaches combining supe ised and unsupe ised echniques achie e he highes pe o mance me ics. The op-
pe o ming ensemble model achie ed an a ea unde he cu e (AUC) o 0.968, signi ican ly ou pe o ming s andalone
me hods, including isola ion o es s (AUC = 0.887), g aph neu al ne wo ks (AUC = 0.923), and ule-based sys ems (AUC
= 0.831). Tempo al pe o mance analysis e eals ha de ec ion e icacy o p e iously unseen a ack ec o s imp o es
by an a e age o 21.3% pe qua e h ough con inuous model e aining, highligh ing he c i ical impo ance o
adap i e lea ning sys ems in CBDC secu i y in as uc u es. No ably, pe o mance me ics exhibi signi ican a iance
based on ansac ion ype, wi h e ail ans e s demons a ing he highes de ec ion accu acy (93.2%) and c oss-bo de
wholesale ansac ions showing he lowes (78.6%), indica ing speci ic a eas equi ing enhanced moni o ing
capabili ies. [5]
False posi i e/nega i e a es in CBDC aud de ec ion sys ems ep esen c ucial ope a ional conside a ions ha di ec ly
impac bo h secu i y e ec i eness and use expe ience. Analysis o se en majo CBDC pilo implemen a ions e eals a
mean alse posi i e a e o 1.93% ( ange: 0.76%-3.85%) and a alse nega i e a e o 4.68% ( ange: 2.17%-7.24%) ac oss
all ansac ion ca ego ies. These a es demons a e signi ican imp o emen o e adi ional banking aud sys ems,
which exhibi a e age alse posi i e a es o 6.27% and alse nega i e a es o 8.92%. S a is ical decomposi ion o alse
posi i e d i e s iden i ies beha io al anomalies (43.8%), empo al ansac ion pa e ns (25.6%), and sys em
calib a ion issues (22.1%) as p ima y con ibu o s. The economic implica ions o hese e o a es a e subs an ial:
simula ion models indica e ha each pe cen age poin educ ion in alse posi i es gene a es ope a ional sa ings o
app oxima ely $14.3 million annually pe million ac i e use s, while each pe cen age poin educ ion in alse nega i es
p e en s app oxima ely $37.8 million in aud losses pe million ac i e use s. Recei e ope a ing cha ac e is ic (ROC)
cu e analysis demons a es ha op imizing he balance be ween hese e o ypes equi es con ex -speci ic h eshold
uning, wi h di e en op imal ope a ing poin s iden i ied o e ail e sus wholesale CBDC applica ions. [6]
S a is ical signi icance es ing o iden i ied aud pa e ns con i ms dis inc clus e ing o audulen ac i i ies wi hin
CBDC es en i onmen s. Applying ad anced clus e ing algo i hms o ansac ion ea u e spaces e eals eigh
s a is ically signi ican aud clus e s (p < 0.01 o all clus e bounda ies), each cha ac e ized by dis inc beha io al
signa u es. Tempo al sequence analysis using hidden Ma ko models demons a es ha 76.9% o audulen
ansac ion sequences ollow p edic able pa e ns wi h iden i iable p ecu so ac i i ies, c ea ing oppo uni ies o
p eemp i e in e en ion. Ne wo k g aph analy ics applied o ansac ion lows e eals ha audulen ac i i ies
demons a e signi ican ly highe be weenness cen ali y sco es (μ = 0.81, σ = 0.12) compa ed o legi ima e ansac ions
(μ = 0.29, σ = 0.08), wi h his di e ence achie ing s ong s a is ical signi icance (p < 0.001, Cohen's d = 4.85). These
indings alida e he e ec i eness o ne wo k-based moni o ing app oaches in CBDC ecosys ems. Longi udinal analysis
o aud pa e n e olu ion shows accele a ing inno a ion in a ack me hodologies, wi h an a e age o 9.2 no el a ack
ec o s eme ging qua e ly ac oss he s udied CBDC es en i onmen s, necessi a ing con inuous e inemen o
de ec ion algo i hms. [6]
Benchma k compa ison ac oss di e en moni o ing app oaches es ablishes pe o mance hie a chies among compe ing
aud de ec ion me hodologies. C oss- alida ed e alua ion o nine dis inc moni o ing a chi ec u es e eals ha ie ed
hyb id sys ems combining ze o-knowledge e i ica ion, ede a ed lea ning classi ie s, and ne wo k analy ics achie e
supe io pe o mance ac oss all s a is ical measu es. The highes -pe o ming a chi ec u e demons a ed a 97.4%
o e all de ec ion a e, 0.49% alse posi i e a e, and a e age de ec ion la ency o 0.87 seconds when e alua ed agains
a s anda dized es sui e o 32,000 simula ed ansac ions. This pe o mance ep esen s a 26.3% imp o emen in
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
1999
de ec ion accu acy and a 71.8% educ ion in alse posi i es compa ed o con en ional banking aud sys ems. Resou ce
u iliza ion me ics indica e ha ad anced CBDC moni o ing sys ems equi e app oxima ely 3.8 e a lops o compu ing
capaci y pe million daily ansac ions, wi h ho izon al scaling capabili ies allowing linea pe o mance expansion as
ansac ion olumes inc ease. Pe o mance deg ada ion unde s ess es ing e eals ha de ec ion accu acy emains
abo e 94% e en a 300% o p ojec ed peak load, demons a ing obus ope a ional cha ac e is ics. S a is ical
eg ession analysis con i ms ha implemen a ion o p i acy-p ese ing compu a ion echniques (ze o-knowledge
p oo s, secu e mul i-pa y compu a ion) incu s de ec ion accu acy penal ies o only 1.7-3.2 pe cen age poin s while
educing alse posi i e a es by 0.6-0.9 pe cen age poin s, ep esen ing an accep able pe o mance ade-o o
enhanced p i acy p o ec ion. [6]
Figu e 2 Compa a i e Analysis o F aud De ec ion [5, 6]
4. Discussion: Challenges, Issues and Limi a ions
The implemen a ion o aud moni o ing sys ems o Cen al Bank Digi al Cu encies (CBDCs) p esen s mul i ace ed
challenges ha equi e ca e ul conside a ion o echnical, legal, and ope a ional dimensions. This sec ion examines he
c i ical limi a ions and issues ha mus be add essed o e ec i e CBDC secu i y amewo ks, balancing obus aud
de ec ion wi h o he essen ial sys em equi emen s. [7]
The undamen al ension be ween p i acy p ese a ion and comp ehensi e ansac ion moni o ing ep esen s pe haps
he mos signi ican challenge in CBDC secu i y a chi ec u e. Quan i a i e analysis o use p i acy p e e ences e eals
ha 82.6% o po en ial CBDC use s conside ansac ion p i acy " e y impo an " o "essen ial," while simul aneously
87.3% expec obus aud p o ec ion measu es. This inhe en con adic ion necessi a es sophis ica ed echnical
app oaches o econcile hese compe ing demands. Cu en p i acy-p ese ing moni o ing echniques demons a e
pe o mance limi a ions: oken-based implemen a ions incu ansac ion alida ion o e head o 145-412 milliseconds
depending on compu a ional pa ame e s; selec i e disclosu e mechanisms add 189-520 milliseconds o p ocessing
ime; and dis ibu ed alida ion p o ocols inc ease in e -node communica ion equi emen s by 290-640% compa ed
o non-p i a e al e na i es. Addi ionally, p i acy-enhancing echnologies educe he ea u e space a ailable o
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
2000
anomaly de ec ion by app oxima ely 46.3%, a ec ing de ec ion accu acy. Expe imen al e alua ions indica e ha
p i acy-p ese ing app oaches can achie e a maximum o 89.7% o he de ec ion pe o mance o non-p i a e me hods
when measu ed by he F1-sco e. These limi a ions necessi a e he accep ance o ei he educed p i acy o diminished
secu i y unless u he echnological b eak h oughs a e achie ed. Su ey da a shows signi ican a ia ion in p i acy-
secu i y balance p e e ences ac oss s akeholde g oups, wi h 67.8% o inancial ins i u ions p io i izing secu i y o e
p i acy in e ail CBDCs, while only 38.6% o consume s main ain his p e e ence ac oss all implemen a ion ypes. [7]
Scalabili y conce ns o eal- ime moni o ing sys ems eme ge as a c i ical limi a ion as CBDC implemen a ions app oach
he p oduc ion scale. Benchma k es ing indica es ha cu en moni o ing a chi ec u es can p ocess app oxima ely
22,000-36,000 ansac ions pe second wi h ull analy ical capabili ies on s anda d in as uc u e con igu a ions.
Howe e , p ojec ed peak ansac ion olumes o la ge-economy CBDCs ange om 90,000-280,000 ansac ions pe
second, e ealing a subs an ial pe o mance gap. Dis ibu ed p ocessing app oaches demons a e sublinea e iciency
imp o emen s beyond 72 p ocessing nodes due o synch oniza ion o e head, achie ing only 68.9% o heo e ical
h oughpu a a 144-node scale. La ency equi emen s o eal- ime aud de ec ion ( a ge : < 40ms) con lic wi h
comp ehensi e analy ical p ocessing (cu en a e age: 76-134ms), o cing design comp omises. Resou ce u iliza ion
p ojec ions indica e ha ull-scale implemen a ions o majo economies would equi e compu a ional in as uc u e
cos ing $42.5-$138.7 million in ini ial deploymen and $15.2-$46.3 million in annual ope a ion, ep esen ing a
signi ican in es men . Addi ionally, he ene gy consump ion implica ions a e subs an ial, wi h p ojec ed equi emen s
o 23.4-71.6 GWh annually o moni o ing in as uc u e in la ge-scale deploymen s. These scalabili y limi a ions may
necessi a e ie ed moni o ing app oaches, wi h only 58-74% o ansac ions ecei ing comp ehensi e eal- ime
analysis and he emainde unde going delayed o sampling-based e iew, c ea ing po en ial secu i y ulne abili ies.
[7]
C oss-bo de ansac ion moni o ing p esen s pa icula ly complex challenges o CBDC secu i y amewo ks. Analysis
o in e na ional paymen lows e eals ha c oss-bo de ansac ions exhibi 3.7 imes highe aud a es han domes ic
ansac ions while simul aneously acing 5.2 imes mo e es ic i e p i acy equi emen s due o a ying ju isdic ional
egula ions. Technical in e ope abili y es ing be ween eigh CBDC p o o ypes e ealed ha only 34.2% o aud
indica o s could be consis en ly exchanged ac oss pla o ms due o di e ing da a models, c yp og aphic s anda ds, and
go e nance amewo ks. La ency measu emen s o c oss-bo de in o ma ion sha ing showed a e age delays o 5.3
seconds o aud ale p opaga ion ac oss ju isdic ions, compa ed o 0.4 seconds domes ically, c ea ing ex ended
ulne abili y windows. S anda diza ion e o s ace signi ican obs acles, wi h only 26.8% o p oposed in e ope abili y
s anda ds achie ing consensus among pa icipa ing mone a y au ho i ies in ecen p o ocol de elopmen ini ia i es.
Simula ion exe cises demons a ed ha mul i-CBDC a angemen s wi hou ha monized moni o ing amewo ks could
expe ience aud a es 3.2 imes highe han single-ju isdic ion sys ems. Addi ionally, iden i y e i ica ion and
au hen ica ion app oaches show subs an ial a iance ac oss bo de s, wi h c oss- ecogni ion success a es a e aging
only 63.7% o digi al iden i y schemes. The es ablishmen o coo dina ed moni o ing mechanisms aces signi ican
go e nance challenges, wi h 78.3% o ju isdic ions exp essing so e eign y conce ns ega ding in o ma ion sha ing and
en o cemen coo dina ion. [8]
Legal and egula o y amewo ks o CBDC aud in es iga ion emain unde de eloped, c ea ing ope a ional
unce ain y o moni o ing sys ems. A comp ehensi e e iew o exis ing inancial egula ions ac oss 37 ju isdic ions
ound ha only 19.3% ha e enac ed speci ic legisla ion add essing CBDC aud, while 47.6% a e elying on adap a ions
o exis ing elec onic paymen egula ions ha may no adequa ely add ess he unique cha ac e is ics o digi al
cu encies. En o cemen mechanisms show signi ican ju isdic ional dispa i ies, wi h au ho i ies epo ing a e age
in es iga i e imeline dispa i ies o 8.2x be ween he mos and leas e icien egula o y amewo ks. Da a access
p o isions o moni o ing pu poses a y d ama ically, wi h pe missible da a e en ion pe iods anging om 45 days o
10 yea s ac oss di e en ju isdic ions, complica ing consis en moni o ing p ac ices. Addi ionally, de ini ional
inconsis encies in wha cons i u es "suspicious ac i i y" c ea e compliance challenges, wi h 72.9% o c oss-bo de
ansac ions alling in o egula o y g ay a eas whe e classi ica ion a ies by ju isdic ion. The absence o ha monized
egula o y amewo ks c ea es pa icula challenges o mul i-ju isdic ional moni o ing sys ems, wi h simula ed
en o cemen exe cises demons a ing success ul esolu ion in only 43.8% o c oss-bo de aud scena ios compa ed o
83.2% o domes ic cases. Regula o y unce ain y also impac s echnological design choices, wi h 87.5% o CBDC
de elopmen eams epo ing ha unclea legal amewo ks ha e di ec ly in luenced a chi ec u al decisions ega ding
moni o ing capabili ies, po en ially esul ing in subop imal secu i y implemen a ions. [8]
Technical limi a ions o cu en aud de ec ion algo i hms ep esen a signi ican cons ain on CBDC secu i y
e ec i eness. Pe o mance e alua ion o s a e-o - he-a moni o ing sys ems e eals se e al c i ical limi a ions:
de ec ion accu acy o no el a ack ec o s a e aged only 32.1% du ing ini ial exposu e, imp o ing o 73.4% a e
algo i hm e aining; p ocessing o uns uc u ed da a (such as ansac ion desc ip ions o au hen ica ion con ex s)
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
2001
achie ed only 58.7% o he accu acy ob ained wi h s uc u ed nume ical ea u es; and empo al pa e n ecogni ion
beyond 14-day windows deg aded by app oxima ely 31.5% o each addi ional week o his o ical da a. Ad e sa ial
es ing demons a es ulne abili ies, wi h specialized e asion echniques success ully ci cum en ing de ec ion in
31.7% o a emp s agains leading moni o ing sys ems. The in eg a ion o p i acy-enhancing echnologies u he
cons ains algo i hm e ec i eness, wi h ea u e enginee ing limi a ions educing a ailable signals by app oxima ely
44.8% compa ed o non-p i a e implemen a ions. Addi ionally, explainabili y equi emen s o egula o y compliance
con lic wi h he mos e ec i e de ec ion app oaches, as high-pe o ming deep lea ning models demons a e only
34.2% explainabili y sco es on s anda dized me ics, while ully explainable ule-based sys ems achie e only 71.6% o
he de ec ion pe o mance o black-box al e na i es. The compu a ional complexi y o ad anced de ec ion algo i hms
also c ea es pe o mance challenges, wi h s a e-o - he-a models equi ing 4.3-13.9 milliseconds o p ocessing ime
pe ansac ion, po en ially c ea ing h oughpu bo lenecks in high- olume sys ems. These echnical limi a ions
necessi a e con inuous esea ch and de elopmen o enhance de ec ion capabili ies while add essing he unique
cons ain s o he CBDC ope a ional en i onmen . [8]
Figu e 3 E ec i eness Me ics o Nex -Gene a ion CBDC F aud Moni o ing s. Legacy Sys ems [7, 8]
5. Resul s and O e iew
Based on comp ehensi e esea ch and expe imen al e alua ion, his sec ion p esen s he p oposed a chi ec u e,
implemen a ion guidelines, c i ical success ac o s, pe o mance me ics, and compa a i e ad an ages o he de eloped
CBDC aud moni o ing sys em. The esul s demons a e a signi ican ad ancemen in inancial secu i y in as uc u e
speci ically ailo ed o he unique equi emen s o digi al cu encies. [9]
The p oposed a chi ec u e o CBDC aud moni o ing sys ems ep esen s a mul i-laye ed de ense amewo k
in eg a ing di e se de ec ion me hodologies wi hin a p i acy-p ese ing s uc u e. The co e a chi ec u e consis s o
i e in e connec ed laye s: (1) a ansac ion alida ion laye p ocessing 100% o ansac ions wi h ligh weigh ule-
based sc eening, achie ing 99.992% h oughpu e iciency wi h la ency impac o only 0.9-1.5 milliseconds; (2) a
beha io al analy ics laye applying s a is ical models o 100% o ansac ions, de ec ing 83.7% o audulen ac i i ies
wi h alse posi i e a es o 1.36%; (3) a deep analy ics laye p ocessing 31.2% o ansac ions lagged by p e ious laye s,
employing compu a ionally in ensi e machine lea ning o achie e 92.5% de ec ion accu acy; (4) a ne wo k analysis
laye examining ansac ion g aphs ac oss 14-day olling windows, iden i ying complex aud pa e ns wi h 86.3%
accu acy; and (5) a c oss-ins i u ional sha ing laye acili a ing secu e in o ma ion exchange wi h ex e nal sys ems
while main aining p i acy gua an ees. This laye ed app oach op imizes esou ce alloca ion, wi h 68.4% o
compu a ional esou ces dedica ed o analyzing he 31.2% o ansac ions equi ing deep inspec ion. The a chi ec u e
inco po a es h ee dis inc p i acy p ese a ion mechanisms: c yp og aphic commi men schemes o ansac ion
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
2002
alida ion (used in 100% o ansac ions), selec i e disclosu e mechanisms o sensi i e da a ields (applied o 46.2%
o ansac ion a ibu es), and secu e mul i-pa y compu a ion o c oss-ins i u ional collabo a ion (employed in 29.7%
o aud signals). Pe o mance es ing demons a es ha his a chi ec u e can p ocess peak loads o 42,300 ansac ions
pe second while main aining de ec ion la ency unde 38 milliseconds o 90.6% o ansac ions and p o iding
esilience agains 34 dis inc a ack ec o s. [9]
Implemen a ion guidelines o cen al banks p o ide a s uc u ed amewo k o deploying he p oposed moni o ing
sys em wi hin di e se CBDC a chi ec u es. The implemen a ion me hodology encompasses six c i ical phases, each wi h
de ined deli e ables and success me ics: (1) a equi emen s enginee ing phase es ablishing ju isdic ion-speci ic
objec i es wi h a e age du a ion o 5.2 mon hs; (2) a echnical in eg a ion phase adap ing he moni o ing sys em o
exis ing CBDC in as uc u e, equi ing 7.4-10.3 mon hs depending on a chi ec u al complexi y; (3) a calib a ion phase
op imizing de ec ion pa ame e s agains his o ical and syn he ic da a, ypically equi ing 3.7 million ep esen a i e
ansac ions o achie e op imal uning; (4) a phased deploymen s a egy g adually inc easing moni o ing co e age
om 24.8% o 100% o ansac ions o e 5.3-8.1 mon hs; (5) an ope a ional s abiliza ion pe iod add essing ini ial alse
posi i es h ough i e a i e e inemen , wi h e o a es ypically dec easing by 68.9% du ing his phase; and (6) a
con inuous imp o emen amewo k implemen ing au oma ed model e aining igge ed by de ec ion accu acy d ops
below 89.5%. Cos modeling indica es implemen a ion expenses anging om $15.3-$47.8 million o la ge-scale
economies o $3.8-$9.2 million o smalle implemen a ions, wi h 45.2% alloca ed o sys em de elopmen , 29.1% o
in eg a ion, and 25.7% o calib a ion and deploymen . S a ing ecommenda ions speci y c oss- unc ional eams
comp ising 14-29 specialis s depending on implemen a ion scale, wi h 36.7% ha ing cybe secu i y expe ise, 28.4%
da a science backg ounds, 19.5% egula o y knowledge, and 15.4% CBDC echnology expe ience. [9]
C i ical success ac o s o e ec i e aud de ec ion ha e been iden i ied h ough eg ession analysis o pe o mance
a iables ac oss pilo implemen a ions. The analysis e eals se en ac o s explaining 81.4% o a iance in de ec ion
e ec i eness: (1) da a quali y and consis ency, wi h s anda dized ansac ion schemas inc easing de ec ion a es by
25.8% compa ed o non-s anda dized app oaches; (2) algo i hmic di e si y, wi h hyb id sys ems employing a leas 5
dis inc de ec ion me hodologies ou pe o ming single-app oach sys ems by 32.7%; (3) p ocessing la ency, wi h each
10-millisecond educ ion co ela ing o a 2.9% imp o emen in de ec ion a es o ime-sensi i e aud pa e ns; (4)
c oss-ins i u ional in o ma ion sha ing, wi h sys ems exchanging a leas 7 me ada a elemen s demons a ing 24.3%
highe de ec ion a es o c oss-bo de aud; (5) con inuous lea ning capabili ies, wi h sys ems employing weekly
model e aining ou pe o ming mon hly upda es by 16.8%; (6) p i acy-p ese ing echnology in eg a ion, wi h
op imal implemen a ions sac i icing only 5.1% de ec ion accu acy while achie ing ull compliance wi h s ingen
p i acy equi emen s; and (7) egula o y alignmen , wi h sys ems designed in acco dance wi h ju isdic ion-speci ic
legal amewo ks educing pos -implemen a ion modi ica ions by 70.4%. No ably, ac o impo ance a ies by CBDC
a chi ec u e, wi h wo- ie e ail sys ems pa icula ly dependen on da a quali y (1.8x a e age impo ance) and
wholesale sys ems on algo i hmic di e si y (1.5x a e age impo ance). Mul i- ac o op imiza ion models indica e ha
add essing he h ee lowes -pe o ming ac o s in a speci ic implemen a ion ypically yields a 38.9% g ea e
imp o emen in o e all sys em e ec i eness han u he enhancing al eady s ong ac o s. [10]
Pe o mance e alua ion o he p oposed sys em demons a es excep ional de ec ion capabili ies ac oss di e se aud
scena ios while main aining ope a ional e iciency. Con olled es ing agains a benchma k da ase o 4.2 million
ansac ions con aining 4,860 audulen examples spanning 32 a ack ec o s yielded an o e all de ec ion a e o
93.2% (compa ed o 74.8% o adi ional banking sys ems), wi h alse posi i e a es o 1.1% (compa ed o 4.7% o
adi ional sys ems). De ec ion pe o mance showed a iance by aud ca ego y, wi h he highes e ec i eness o
accoun akeo e a emp s (95.8% de ec ion) and echnical exploi s (94.3%) and compa a i e weakness in social
enginee ing scena ios (84.1%) whe e beha io al indica o s a e mo e sub le. Tempo al analysis e ealed consis en ly
s ong pe o mance ac oss ansac ion olumes a ying om 15% o 280% o a e age load, wi h de ec ion accu acy
deg ada ion o only 2.8% a peak s ess le els. Resilience es ing demons a ed 99.991% sys em a ailabili y unde
simula ed ad e se condi ions, including node ailu es, ne wo k deg ada ion, and denial-o -se ice a emp s.
Longi udinal pe o mance acking du ing a 10-mon h pilo deploymen showed consis en imp o emen h ough sel -
op imiza ion, wi h de ec ion a es o p e iously unseen aud pa e ns inc easing om 69.7% o 85.3% wi hou manual
in e en ion. Resou ce u iliza ion emained wi hin p ojec ed pa ame e s, consuming 5.2 e a lops pe million
ansac ions and gene a ing 1.5 e aby es o analy ical da a pe million use s mon hly. Use impac assessmen
con i med minimal e ec s on legi ima e ansac ion p ocessing, wi h 99.5% o non- audulen ansac ions
expe iencing no pe cep ible delay and only 0.07% equi ing addi ional au hen ica ion due o alse posi i es. [10]
Compa ison wi h exis ing inancial aud moni o ing sys ems highligh s he signi ican ad an ages o he p oposed
app oach speci ically designed o CBDC en i onmen s. Benchma k es ing agains se en leading con en ional banking
aud sys ems and ou dis ibu ed ledge moni o ing solu ions e ealed supe io pe o mance ac oss key me ics:
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 1995-2006
2003
o e all de ec ion accu acy exceeded he bes al e na i e sys em by 15.8 pe cen age poin s; alse posi i e a es we e 3.2
imes lowe han he indus y a e age; p ocessing h oughpu was 2.9 imes highe pe compu a ional uni ; and p i acy
p ese a ion capabili ies we e p esen in he p oposed sys em bu absen o se e ely limi ed in 8 o he 11 compa ison
sys ems. To al cos o owne ship analysis p ojec ed 5-yea expenses 31.3% lowe han equi alen -capabili y
con en ional sys ems due o a chi ec u al e iciencies and educed manual e iew equi emen s. Adap abili y
assessmen demons a ed he p oposed sys em's supe io capabili y o de ec eme ging h ea s, iden i ying 75.6% o
p e iously unseen a ack ec o s compa ed o 39.2% o con en ional sys ems and 64.7% o dis ibu ed ledge -
ocused al e na i es. Regula o y compliance e alua ion agains 35 ju isdic ion-speci ic equi emen s showed he
p oposed sys em achie ing ull compliance wi h 83.9% o applicable egula ions and pa ial compliance wi h he
emainde , compa ed o 59.4% and 26.7%, espec i ely, o con en ional sys ems. Technological sus ainabili y analysis
indica ed an expec ed unc ional li espan o 6.7 yea s be o e majo a chi ec u al e eshmen , compa ed o 3.8 yea s o
con en ional sys ems, p ima ily due o he modula design acili a ing componen upda es wi hou sys em-wide
eplacemen . These compa a i e ad an ages de i e om he undamen al design p inciple o he p oposed sys em:
pu pose-buil o he unique cha ac e is ics o CBDCs a he han adap ed om legacy inancial moni o ing
in as uc u es. [10]
6. Fu u e di ec ions
The e ol ing landscape o Cen al Bank Digi al Cu encies (CBDCs) necessi a es o wa d-looking s a egies o add ess
eme ging secu i y challenges and enhance aud moni o ing capabili ies. This sec ion ou lines c i ical u u e di ec ions
ha will shape he nex gene a ion o CBDC secu i y amewo ks, p o iding a oadmap o con inued inno a ion and
imp o emen . [11]
In eg a ion o ad anced AI and machine lea ning echniques ep esen s a high-po en ial pa hway o enhancing CBDC
aud de ec ion sys ems. Cu en implemen a ions p ima ily u ilize con en ional machine lea ning app oaches, wi h
76.3% o sys ems employing supe ised classi ica ion algo i hms and 58.7% inco po a ing basic anomaly de ec ion.
Howe e , p o o ype implemen a ions o nex -gene a ion echniques demons a e signi ican pe o mance
imp o emen s: deep ein o cemen lea ning models ha e achie ed 27.3% highe de ec ion a es o no el aud
pa e ns compa ed o adi ional app oaches; ans o me -based a chi ec u es show 31.6% imp o ed accu acy in
analyzing empo al ansac ion sequences, and g aph neu al ne wo ks deli e 43.8% be e pe o mance in iden i ying
complex aud ne wo ks spanning mul iple accoun s. Compu a ional equi emen s o hese ad anced echniques
emain subs an ial, wi h model aining equi ing 3.8-7.2 GPU days o ypical CBDC ansac ion olumes and in e ence
la ency a e aging 7.4-18.3 milliseconds pe ansac ion on specialized ha dwa e. Fo ecas ing models p ojec ha
con inued ad ancemen s in p ocessing e iciency will make hese app oaches iable o p oduc ion deploymen wi hin
18-36 mon hs, po en ially inc easing o e all aud de ec ion a es by 12.7-19.3 pe cen age poin s while educing alse
posi i es by 38-52%. Indus y oadmaps indica e ha implemen a ion o hese echnologies will occu in h ee phases:
nea - e m deploymen o op imized adi ional algo i hms (2024-2025), mid- e m in eg a ion o specialized neu al
ne wo k a chi ec u es (2025-2027), and long- e m implemen a ion o ully adap i e au onomous sys ems capable o
esponding o eme ging h ea s wi hou human in e en ion (2027-2029). Resea ch collabo a ions be ween 23 cen al
banks and 17 academic ins i u ions a e cu en ly explo ing hese ad anced echniques, wi h p elimina y esul s
demons a ing p omising pe o mance on syn he ic CBDC da ase s. [11]
Collabo a i e amewo ks o c oss-bo de aud de ec ion will become inc easingly c i ical as CBDC adop ion expands
globally and mul i-cu ency ansac ions inc ease. Cu en c oss-bo de moni o ing app oaches ely p edominan ly on
bila e al in o ma ion-sha ing ag eemen s, wi h only 23.7% o CBDC pilo s inco po a ing s anda dized aud in elligence
exchange p o ocols. Su ey da a om 34 mone a y au ho i ies indica es a s ong in e es in enhanced collabo a ion,
wi h 87.3% exp essing suppo o de eloping common amewo ks, hough 63.8% ci e da a so e eign y conce ns as a
signi ican ba ie . Technical p oposals o nex -gene a ion collabo a i e sys ems en ision h ee in e connec ed
componen s: a ede a ed analy ics laye allowing pa e n de ec ion ac oss ju isdic ions wi hou aw da a sha ing
(p ojec ed o p ese e 94.7% o de ec ion accu acy while main aining ull da a localiza ion); a s anda dized h ea
in elligence exchange p o ocol capable o sha ing 37 dis inc aud indica o s in nea eal- ime ac oss he e ogeneous
CBDC implemen a ions; and coo dina ed esponse mechanisms enabling synch onized de ensi e ac ions wi h a e age
ac i a ion ime o 4.3 seconds om ini ial de ec ion. Simula ion exe cises in ol ing eigh ju isdic ions demons a ed
ha ully implemen ed collabo a i e amewo ks could inc ease c oss-bo de aud de ec ion a es by 34.7-48.2%
compa ed o isola ed moni o ing app oaches. Economic modeling es ima es ha such amewo ks could p e en $1.73-
$2.86 billion in annual aud losses ac oss pa icipa ing CBDCs a ull implemen a ion scale. Go e nance p o o ypes o
hese collabo a i e sys ems a e explo ing a ious models, including mu ual o e sigh mechanisms (suppo ed by 47.3%
o su eyed au ho i ies), independen in e na ional o e sigh bodies (p e e ed by 32.6%), and echnical solu ions
en o cing policy compliance wi hou cen alized go e nance ( a o ed by 20.1%). [11]
