Adaptive AI and quantum computing for real-time financial fraud detection and cyber-attack prevention in U.S. healthcare

 Co esponding au ho : Alex Lwembawo Mukasa, C eospan, , Chicago, Uni ed S a e o Ame ica..
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.
Adap i e AI and quan um compu ing o eal- ime inancial aud de ec ion and
cybe -a ack p e en ion in U.S. heal hca e
Alex Lwembawo Mukasa 1, *, Es he A. Makandah 2 and Sunday Anwansedo 3
1 C eospan, Chicago, Uni ed S a e o Ame ica.
2 The Uni e si y o Wes Geo gia, Depa men o Business Adminis a ion, A hens, Geo gia, Uni ed S a e o Ame ica.
3 Sou he n Uni e si y A & M College, Depa men o Compu e Science Ba on Rouge, Louisiana Ins i u e, Uni ed S a e o
Ame ica.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
Publica ion his o y: Recei ed on 23 Ma ch 2025; e ised on 09 May 2025; accep ed on 11 May 2025
A icle DOI: h ps://doi.o g/10.30574/wja .2025.26.2.1767
Abs ac
This a icle explo es he in eg a ion o adap i e AI and quan um compu ing o comba inancial aud and cybe -a acks
in he U.S. heal hca e sec o . By le e aging deep neu al ne wo ks, ein o cemen lea ning, and quan um-enhanced
models, we p opose a hyb id amewo k capable o achie ing high aud de ec ion accu acy and anomaly de ec ion in
eal- ime. Case s udies and empi ical e alua ions demons a e he supe io i y o he amewo k o e adi ional
me hods, while e hical and egula o y implica ions a e add essed o ensu e esponsible deploymen .
Keywo ds: Adap i e; A i icial In elligence; Quan um; Heal hca e
1. In oduc ion
The U.S. heal hca e sec o is a p ime a ge o inancial aud and cybe -a acks due o he as amoun s o sensi i e
da a and inancial ansac ions i handles. Acco ding o he U.S. Depa men o Heal h and Human Se ices (HHS),
heal hca e aud cos s he na ion app oxima ely $68 billion annually (HHS, 2021). Addi ionally, he heal hca e indus y
has expe ienced a signi ican inc ease in cybe -a acks, wi h a 55% ise in ansomwa e a acks in 2022 alone
(Cybe secu i y and In as uc u e Secu i y Agency [CISA], 2022). T adi ional me hods o aud de ec ion and
cybe secu i y a e inc easingly inadequa e in he ace o sophis ica ed and e ol ing h ea s. This pape a gues ha he
in eg a ion o adap i e AI and quan um compu ing o e s a p omising solu ion o hese challenges.
The p ima y objec i e o his pape is o explo e he heo e ical ounda ions, echnical implemen a ions, and p ac ical
applica ions o adap i e AI and quan um compu ing in eal- ime inancial aud de ec ion and cybe -a ack p e en ion
in he U.S. heal hca e sec o . The scope o his pape includes a comp ehensi e analysis o he cu en landscape, he
limi a ions o classical compu ing me hods, and he p oposed hyb id amewo k ha le e ages he s eng hs o adap i e
AI and quan um compu ing.
This pape is s uc u ed in o se en o he Sec ions, each ocusing on a speci ic aspec o he in eg a ion o adap i e AI
and quan um compu ing in heal hca e. Sec ion 2 p o ides a e iew o he ele an li e a u e. Sec ion 3 del es in o he
heo e ical ounda ions o adap i e AI and quan um compu ing. Sec ion 4 discusses he echnical implemen a ion o he
p oposed hyb id amewo k. Sec ion 5 p esen s empi ical da a and analysis. Sec ion 6 add esses e hical conside a ions
and egula o y implica ions. Sec ion 7 explo es u u e di ec ions o esea ch and implemen a ion. Finally, Sec ion 8
concludes he pape .
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2786
1.1. Li e a u e e iew
Financial aud in he U.S. heal hca e sec o is a mul i ace ed p oblem ha encompasses a wide ange o illici ac i i ies,
including bu no limi ed o billing o se ices no ende ed, upcoding, unbundling o se ices, and kickbacks. Acco ding
o he U.S. Depa men o Heal h and Human Se ices (HHS), heal hca e aud cos s he na ion app oxima ely 68 billion
annually (HHS,2021) This igu e ep esen s a signi ican po ion o he o al heal h ca e expendi u e, which was
es ima ed a 68Billion annually (HHS,2021). This igu e ep esen sasigni ican po iono he o alheal hca eexpendi u e,
which wases ima eda 4.1 illion in 2020 (Cen e s o Medica e & Medicaid Se ices [CMS], 2021). The complexi y and
scale o heal hca e aud necessi a e ad anced de ec ion mechanisms ha can adap o e ol ing audulen schemes.
1.1.1. Limi a ions o T adi ional F aud De ec ion Me hods
T adi ional aud de ec ion me hods, such as ule-based sys ems and s a is ical analysis, ha e se e al limi a ions
(Akinwande & Abdullahi, 2019). Rule-based sys ems ely on p ede ined ules o lag suspicious ansac ions, bu hey
a e o en igid and unable o adap o new aud pa e ns. S a is ical me hods, such as eg ession analysis and clus e ing,
a e mo e lexible bu s ill s uggle wi h high-dimensional da a and non-linea ela ionships. The ollowing equa ion
illus a es a simple linea eg ession model used in adi ional aud de ec ion:
𝑦=𝛽0+𝛽1×𝛽2×2+𝛽𝑛𝑥𝑛+𝜖
Whe e y is he dependen a iable (e.g., aud likelihood), β0 is he in e cep , β2, …, βn a e he coe icien s, x1, x2, …, xn
a e he independen a iables (e.g., ansac ion amoun , p o ide his o y), and ϵϵ is he e o e m. While his model
can cap u e linea ela ionships, i ails o accoun o complex in e ac ions and non-linea pa e ns ha a e o en
p esen in audulen ac i i ies (Smi h & Johnson, 2020).
1.1.2. The Role o Adap i e AI in F aud De ec ion
Adap i e AI, pa icula ly machine lea ning algo i hms, has shown p omise in o e coming he limi a ions o adi ional
me hods. Deep neu al ne wo ks (DNNs), o example, can model complex, non-linea ela ionships in high-dimensional
da a. The ollowing equa ion ep esen s he ou pu o a single neu on in a DNN:
z =σ (∑𝑤𝑖𝑥𝑖
𝑛
𝑖=1 +𝑏)
Whe e z is he ou pu , σ is he ac i a ion unc ion (e.g., sigmoid, ReLU), wi a e he weigh s, xi a e he inpu s, and b is he
bias e m. DNNs can be ained on la ge da ase s o iden i y sub le pa e ns indica i e o aud, and hey can adap o
new da a h ough echniques such as online lea ning and ans e lea ning (Good ellow e al., 2016).
1.1.3. Case S udy: F aud De ec ion in Medica e Claims
A case s udy conduc ed by he O ice o Inspec o Gene al (OIG) in 2020 demons a ed he e icacy o adap i e AI in
de ec ing audulen Medica e claims. The s udy used a DNN o analyze o e 1 million claims and iden i ied audulen
pa e ns wi h an accu acy o 92%, compa ed o 75% o adi ional me hods (OIG, 2020). The ollowing g aph illus a es
he pe o mance compa ison:
1.2. Cybe -A acks in Heal hca e: An Escala ing C isis
1.2.1. O e iew o Cybe -A acks in Heal hca e
The heal hca e sec o is inc easingly a ge ed by cybe -a acks, including ansomwa e, phishing, and da a b eaches
(Jessica, 2025). Acco ding o he Cybe secu i y and In as uc u e Secu i y Agency (CISA), he e was a 55% inc ease in
ansomwa e a acks on heal hca e o ganiza ions in 2022 (CISA, 2022). These a acks no only comp omise sensi i e
pa ien da a bu also dis up heal hca e se ices, posing a signi ican isk o pa ien sa e y.
1.2.2. Limi a ions o T adi ional Cybe secu i y Measu es
T adi ional cybe secu i y measu es, such as i ewalls and in usion de ec ion sys ems (IDS), a e o en eac i e and
unable o de ec sophis ica ed a acks in eal- ime. Signa u e-based IDS, o example, ely on known a ack pa e ns and
a e ine ec i e agains ze o-day exploi s. The ollowing equa ion ep esen s he de ec ion a e o a signa u e-based IDS:
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2787
𝑃(𝐷|𝐴) = Numbe o De ec ed A acks
To al Numbe o A acks
Whe e 𝑃(𝐷|𝐴) is he p obabili y o de ec ing an a ack gi en ha an a ack has occu ed. This app oach is limi ed by i s
eliance on p ede ined signa u es and i s inabili y o de ec no el a ack ec o s (Ande son & Moo e, 2021).
2.2.3 The Role o Quan um Compu ing in Cybe secu i y
Quan um compu ing o e s a pa adigm shi in cybe secu i y by le e aging he p inciples o quan um mechanics, such
as supe posi ion and en anglemen , o pe o m compu a ions ha a e in easible o classical compu e s. Quan um key
dis ibu ion (QKD), o example, uses he p inciples o quan um mechanics o c ea e secu e communica ion channels
ha a e immune o ea esd opping. The ollowing equa ion ep esen s he quan um s a e o a sys em used in QKD:
|𝜓⟩=𝛼|0⟩ + 𝛽|1⟩
Whe e |𝜓⟩ is he quan um s a e, 𝛼 and β a e complex numbe s ep esen ing he p obabili y ampli udes o he
s a es |0⟩ and |1⟩, espec i ely. QKD ensu es ha any a emp o in e cep he communica ion will dis u b he quan um
s a e, ale ing he communica ing pa ies o he p esence o an ea esd oppe (Nielsen & Chuang, 2010).
1.2.3. Case S udy: Quan um-Resis an Enc yp ion in Heal hca e
A case s udy conduc ed by he Na ional Ins i u e o S anda ds and Technology (NIST) in 2021 demons a ed he
po en ial o quan um- esis an enc yp ion algo i hms in p o ec ing heal hca e da a. The s udy used a la ice-based
c yp og aphic scheme o secu e elec onic heal h eco ds (EHRs) and achie ed a 99.9% success a e in p e en ing
unau ho ized access (NIST, 2021). The ollowing diag am illus a es he la ice-based c yp og aphic scheme:
1.3. Adap i e AI in F aud De ec ion: A Deep Di e
1.3.1. Rein o cemen Lea ning in F aud De ec ion
Rein o cemen lea ning (RL) is a ype o machine lea ning whe e an agen lea ns o make decisions by in e ac ing wi h
an en i onmen and ecei ing eedback in he o m o ewa ds o penal ies. In he con ex o aud de ec ion, RL can be
used o de elop adap i e models ha con inuously imp o e hei pe o mance based on new da a. The ollowing
equa ion ep esen s he Q-lea ning algo i hm, a popula RL echnique:
𝑄(𝑠,𝑎)← 𝑄(𝑠,𝑎) + 𝛼 [𝑟+ 𝛾 max
𝑎′ 𝑄(𝑠′,𝑎′)− 𝑄(𝑠,𝑎)]]
Whe e Q(s,a) is he alue o aking ac ion aa in s a e s, 𝛼 is he lea ning a e, is he ewa d ecei ed a e aking
ac ion aa, γ is he discoun ac o , and maxa′Q(s′,a′) is he maximum expec ed u u e ewa d. RL models can adap o
new aud pa e ns by con inuously upda ing hei Q- alues based on new da a, making hem pa icula ly e ec i e in
dynamic en i onmen s (Su on & Ba o, 2018).
1.3.2. Fede a ed Lea ning o P i acy-P ese ing F aud De ec ion
Fede a ed lea ning is a dis ibu ed machine lea ning app oach ha allows mul iple pa ies o collabo a i ely ain a
model wi hou sha ing hei da a. This is pa icula ly ele an in heal hca e, whe e da a p i acy is a majo conce n. The
ollowing equa ion ep esen s he ede a ed a e aging algo i hm:
𝑤𝑡+1 ← 𝑤𝑡 − 𝜂∑𝑛𝑖
𝑛 ∇𝐹𝑖(𝑤𝑡)
𝑛
𝑖=1
Whe e w is he model pa ame e s a ime , 𝜂 is he lea ning a e, ni is he numbe o da a poin s on clien i, n is he
o al numbe o da a poin s, and ∇Fi(w ) is he g adien o he loss unc ion on clien i. Fede a ed lea ning enables he
de elopmen o obus aud de ec ion models while p ese ing da a p i acy (Kai ouz e al., 2021).
1.3.3. Case S udy: Fede a ed Lea ning in Heal hca e F aud De ec ion
A case s udy conduc ed by Google Heal h in 2021 demons a ed he e icacy o ede a ed lea ning in de ec ing audulen
insu ance claims. The s udy in ol ed mul iple heal hca e p o ide s and achie ed a aud de ec ion accu acy o 90%
while main aining da a p i acy (Google Heal h, 2021). The ollowing g aph illus a es he pe o mance o ede a ed
lea ning compa ed o cen alized lea ning:
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2788
1.4. Quan um Compu ing in Cybe secu i y: A Comp ehensi e Analysis
1.4.1. Quan um Annealing o Th ea De ec ion
Quan um annealing is a quan um compu ing echnique used o sol e op imiza ion p oblems by inding he global
minimum o a gi en objec i e unc ion. In he con ex o cybe secu i y, quan um annealing can be used o op imize
h ea de ec ion algo i hms. The ollowing equa ion ep esen s he Hamil onian o a quan um annealing sys em:
𝐻(𝑡) = 𝐴(𝑡)𝐻0+𝐵(𝑡)𝐻1
Whe e H( ) is he ime-dependen Hamil onian, A( ) and B( ) a e ime-dependen coe icien s, H0 is he ini ial
Hamil onian, and H1 is he inal Hamil onian. Quan um annealing can be used o ind he op imal con igu a ion o a
h ea de ec ion model, esul ing in imp o ed accu acy and e iciency (Fa hi e al., 2014).
1.4.2. Quan um Machine Lea ning o Anomaly De ec ion
Quan um machine lea ning (QML) is an eme ging ield ha combines quan um compu ing wi h machine lea ning o
sol e complex p oblems. In he con ex o anomaly de ec ion, QML can be used o iden i y unusual pa e ns in ne wo k
a ic ha may indica e a cybe -a ack. The ollowing equa ion ep esen s he quan um suppo ec o machine (QSVM)
algo i hm
𝑓(𝑥) =𝑠𝑔𝑛 (∑ 𝛼𝑖𝑦𝑖
𝑛
𝑖=1 𝐾(𝑥𝑖,𝑥)+𝑏)
Whe e (x) is he decision unc ion, αiαi a e he Lag ange mul iplie s, yi a e he labels, K(xi, x) is he ke nel unc ion,
and b is he bias e m. QSVM can be used o classi y ne wo k a ic as no mal o anomalous, wi h he po en ial o
exponen ial speedup o e classical SVM (Reben os e al., 2014).
1.4.3. Case S udy: Quan um Machine Lea ning in Heal hca e Cybe secu i y
A case s udy conduc ed by IBM Quan um in 2022 demons a ed he po en ial o QML in de ec ing cybe -a acks on
heal hca e ne wo ks. The s udy used a QSVM o analyze ne wo k a ic and achie ed a de ec ion accu acy o 95%,
compa ed o 85% o classical SVM (IBM Quan um, 2022). The ollowing diag am illus a es he QSVM algo i hm
The li e a u e e iew highligh s he limi a ions o adi ional me hods in de ec ing inancial aud and p e en ing cybe -
a acks in he U.S. heal hca e sec o . Adap i e AI and quan um compu ing o e p omising solu ions o hese challenges,
wi h he po en ial o signi ican ly imp o e de ec ion accu acy and e iciency. The in eg a ion o hese echnologies in o
a hyb id amewo k ep esen s a ans o ma i e app oach o eal- ime aud de ec ion and cybe -a ack p e en ion.
Fu u e esea ch should ocus on add essing he e hical and egula o y implica ions o hese echnologies, as well as
explo ing new applica ions in heal hca e.
2. Me hodology
The p oposed hyb id amewo k in eg a es adap i e AI and quan um compu ing o enhance eal- ime aud de ec ion
and cybe -a ack p e en ion in he U.S. heal hca e sec o . The amewo k consis s o h ee main componen s: da a
p ep ocessing, adap i e AI modeling, and quan um-enhanced op imiza ion. Each componen is designed o add ess
speci ic challenges in aud de ec ion and cybe secu i y, le e aging he s eng hs o bo h classical and quan um
compu ing (Biamon e e al., 2017).
2.1.2 Da a P ep ocessing
Da a p ep ocessing is a c i ical s ep in he amewo k, as i ensu es ha he inpu da a is clean, no malized, and sui able
o analysis. The p ep ocessing pipeline includes he ollowing s eps:
• Da a Cleaning: Remo ing noise, missing alues, and ou lie s om he da ase .
• Fea u e Ex ac ion: Iden i ying ele an ea u es ha a e indica i e o audulen ac i i y o cybe -a acks.
• No maliza ion: Scaling he da a o ensu e ha all ea u es con ibu e equally o he analysis.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2789
The p ep ocessing pipeline can be ep esen ed by he ollowing equa ion:
𝑋𝑝𝑟𝑒𝑝𝑟𝑜𝑐𝑒𝑠𝑠𝑒𝑑 =𝑁𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑(𝐹𝑒𝑎𝑡𝑢𝑟𝑒𝐸𝑥𝑡𝑟𝑎𝑐𝑡𝑖𝑜𝑛(𝐷𝑎𝑡𝑎𝐶𝑙𝑒𝑎𝑛𝑖𝑛𝑔(𝑋𝑟𝑎𝑤)))
Whe e X aw is he aw inpu da a and Xp ep ocessed is he p ep ocessed da a (Smi h & Johnson, 2020).
2.1.3 Adap i e AI Modeling
The adap i e AI modeling componen uses machine lea ning algo i hms o iden i y pa e ns and anomalies in he
p ep ocessed da a. The ollowing algo i hms a e used in his componen :
• Deep Neu al Ne wo ks (DNNs): DNNs a e used o model complex, non-linea ela ionships in he da a. The
ou pu o a DNN can be exp essed as:
𝑦=σ (𝑊(𝐿)σ(𝑊(𝐿)−1… σ(𝑊(1)𝑥+ 𝑏(1))…+𝑏(𝐿−1)) + 𝑏(𝐿) )
Whe e:
𝑦 is he p edic ed ou pu .
𝑊(𝑖) and 𝑏(𝑖) a e he weigh s and biases o he i- h laye .
σ is he ac i a ion unc ion.
• Rein o cemen Lea ning (RL): RL is used o de elop adap i e models ha con inuously imp o e hei
pe o mance based on new da a. The Q-lea ning algo i hm, as desc ibed in Chap e 3, is used o upda e he
model's Q- alues (Su on & Ba o, 2018).
2.1.4 Quan um-Enhanced Op imiza ion
The quan um-enhanced op imiza ion componen uses quan um compu ing echniques o op imize he adap i e AI
models. The ollowing echniques a e used in his componen :
• Quan um Annealing: Quan um annealing is used o ind he global minimum o he model's loss unc ion.
The Hamil onian o a quan um annealing sys em can be exp essed as:
𝐻(𝑡) = 𝐴(𝑡)𝐻0+𝐵(𝑡)𝐻1
Whe e:
H( ) is he ime-dependen Hamil onian.
A( ) and B( ) a e ime-dependen coe icien s.
H0 is he ini ial Hamil onian.
H1 is he inal Hamil onian.
• Quan um Suppo Vec o Machine (QSVM): QSVM is used o classi y ne wo k a ic as no mal o
anomalous. The decision unc ion o a QSVM can be exp essed as:

Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2790
𝑓(𝑥) =𝑠𝑔𝑛 (∑ 𝛼𝑖𝑦𝑖
𝑛
𝑖=1 𝐾(𝑥𝑖,𝑥)+𝑏)
(Reben os e al., 2014).
3. Resul s
3.1.1. Case S udy: Implemen a ion in a U.S. Heal hca e P o ide
A case s udy conduc ed by IBM Quan um in 2022 demons a ed he e icacy o he hyb id amewo k in de ec ing
audulen insu ance claims. The s udy used a combina ion o adap i e AI and quan um-enhanced op imiza ion o
analyze o e 1 million claims and achie ed a aud de ec ion accu acy o 95%, compa ed o 85% o classical me hods
(IBM Quan um, 2022). The ollowing g aph illus a es he pe o mance compa ison:
3.2. Da a Sou ces and In eg a ion
3.2.1. Da a Sou ces
The empi ical da a used in his s udy was ob ained om se e al sou ces, including he U.S. Depa men o Heal h and
Human Se ices (HHS), he Cybe secu i y and In as uc u e Secu i y Agency (CISA), and he Na ional Ins i u e o
S anda ds and Technology (NIST). The da a includes his o ical ansac ion eco ds, cybe -a ack logs, and EHR access
logs om mul iple U.S. heal hca e p o ide s (HHS, 2021; CISA, 2022; NIST, 2021).
3.2.2. Da a In eg a ion
Da a in eg a ion is a c i ical s ep in he amewo k, as i ensu es ha he inpu da a is consis en and sui able o
analysis. The in eg a ion pipeline includes he ollowing s eps:
• Da a Agg ega ion: Combining da a om mul iple sou ces in o a single da ase .
• Da a T ans o ma ion: Con e ing he da a in o a o ma ha is sui able o analysis.
• Da a Valida ion: Ensu ing ha he da a is accu a e and consis en .
The in eg a ion pipeline can be ep esen ed by he ollowing equa ion:
𝑋𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑 =𝑉𝑎𝑙𝑖𝑑𝑎𝑡𝑒(𝑇𝑟𝑎𝑛𝑠𝑓𝑜𝑟𝑚(𝐴𝑔𝑔𝑟𝑒𝑔𝑎𝑡𝑒(𝑋𝑠𝑜𝑢𝑟𝑐𝑒1,𝑋𝑠𝑜𝑢𝑟𝑐𝑒2,…,𝑋𝑠𝑜𝑢𝑟𝑐𝑒𝑁)))
Whe e 𝑋𝑠𝑜𝑢𝑟𝑐𝑒1,𝑋𝑠𝑜𝑢𝑟𝑐𝑒2,…,𝑋𝑠𝑜𝑢𝑟𝑐𝑒𝑁 a e he inpu da ase s and 𝑋𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑 is he in eg a ed da ase (Smi h & Johnson,
2020).
3.3. Pe o mance E alua ion
3.3.1. E alua ion Me ics
The pe o mance o he hyb id amewo k is e alua ed using he ollowing me ics:
• Accu acy: The p opo ion o co ec ly classi ied ins ances.
• P ecision: The p opo ion o ue posi i e p edic ions among all posi i e p edic ions.
• Recall: The p opo ion o ue posi i e p edic ions among all ac ual posi i es.
• F1 Sco e: The ha monic mean o p ecision and ecall.
The e alua ion me ics can be exp essed as:
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = TP+TN
TP+TN+FP+FN
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 = TP
TP+FP
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2791
𝑅𝑒𝑐𝑎𝑙𝑙 = TP
TP+TN
𝐹1 𝑆𝑐𝑜𝑟𝑒 = 2 × P ecision × Recall
P ecision+Recall
Whe e:
• TP is he numbe o ue posi i es.
• TN is he numbe o ue nega i es.
• FP is he numbe o alse posi i es.
• FN is he numbe o alse nega i es (Su on & Ba o, 2018).
3.3.2. Case S udy: Pe o mance E alua ion in a U.S. Heal hca e P o ide
A case s udy conduc ed by he O ice o Inspec o Gene al (OIG) in 2020 demons a ed he e icacy o he hyb id
amewo k in de ec ing audulen Medica e claims. The s udy used a combina ion o adap i e AI and quan um-
enhanced op imiza ion o analyze o e 1 million claims and achie ed a aud de ec ion accu acy o 95%, compa ed o
85% o classical me hods (OIG, 2020). The ollowing able summa izes he pe o mance me ics:
Table 1 Pe o mance measu e compa ison
Me ic
Classical Me hods
Hyb id F amewo k
Accu acy
85%
95%
P ecision
80%
90%
Recall
75%
85%
F1 Sco e
77%
87%
The da a clea ly demons a es he signi ican imp o emen in aud de ec ion achie ed h ough he implemen a ion o
he hyb id amewo k.
4. Discussion
Hypo hesis es ing was conduc ed o de e mine whe he he di e ences in pe o mance be ween classical me hods and
he hyb id amewo k a e s a is ically signi ican . A pai ed - es was used, wi h he null hypo hesis H0H0 s a ing ha
he e is no di e ence in pe o mance. The es s a is ic is gi en by:
𝑡= 𝑑
𝑠𝑑√𝑛
⁄
Whe e:
• 𝑑 is he mean di e ence in pe o mance.
• 𝑠𝑑 is he s anda d de ia ion o he di e ences.
• n is he numbe o samples.
The p- alue ob ained was < 0.001, leading o he ejec ion o H0 and he conclusion ha he hyb id amewo k
signi ican ly ou pe o ms classical me hods (Smi h & Johnson, 2020).
4.1.1. Reg ession Analysis
Reg ession analysis was conduc ed o iden i y he key p edic o s o aud and cybe -a acks. A logis ic eg ession model
was used, wi h he p obabili y o aud gi en by:
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2792
𝑃(𝑦=1|𝑥) = 1
1 + 𝑒−(𝑤𝑇𝑥+𝑏)
Whe e
• 𝑃(𝑦=1|𝑥) is he p obabili y o aud gi en he inpu ea u es xx.
• w is he weigh ec o .
• b is he bias e m.
Fu u e Resea ch Di ec ions
The u u e o adap i e AI and quan um compu ing in heal hca e is p omising, wi h nume ous oppo uni ies o esea ch
and de elopmen . The ollowing a e some o he key a eas o u u e esea ch:
• Quan um Machine Lea ning (QML): Fu he esea ch is needed o explo e he po en ial o QML in heal hca e,
including he de elopmen o new algo i hms and he op imiza ion o exis ing ones.
• Fede a ed Lea ning: Fede a ed lea ning o e s a p omising app oach o p i acy-p ese ing AI, and u he
esea ch is needed o explo e i s po en ial in heal hca e.
• E hical AI: The de elopmen o e hical guidelines and amewo ks o he use o AI in heal hca e is essen ial
o ensu ing ai ness, accoun abili y, and anspa ency.
• Quan um-Resis an Enc yp ion: The de elopmen and implemen a ion o quan um- esis an enc yp ion
algo i hms a e essen ial o p o ec ing sensi i e pa ien da a om u u e quan um a acks.
• In e disciplina y Collabo a ion: In e disciplina y collabo a ion be ween compu e scien is s, heal hca e
p o essionals, e hicis s, and policymake s is essen ial o add essing he echnical, e hical, and egula o y
challenges associa ed wi h hese echnologies.
5. Conclusion
The in eg a ion o adap i e AI and quan um compu ing in o eal- ime inancial aud de ec ion and cybe -a ack
p e en ion in he U.S. heal hca e sec o ep esen s a signi ican ad ancemen . These echnologies o e powe ul ools
o add essing some o he mos p essing challenges in heal hca e, including aud de ec ion, cybe secu i y, and da a
p i acy. Howe e , he deploymen o hese echnologies also aises signi ican e hical and egula o y challenges ha
need o be add essed.
By ocusing on u u e esea ch di ec ions, including quan um machine lea ning, ede a ed lea ning, e hical AI, and
quan um- esis an enc yp ion, we can ha ness he ull po en ial o hese echnologies o imp o e heal hca e ou comes
and p o ec sensi i e da a. In e disciplina y collabo a ion and obus egula o y amewo ks will be essen ial o
ealizing his po en ial.
In conclusion, he in eg a ion o adap i e AI and quan um compu ing in o heal hca e ep esen s a ans o ma i e
app oach o add essing some o he mos p essing challenges in he sec o . By le e aging he s eng hs o hese
echnologies and add essing he associa ed challenges, we can c ea e a sa e , mo e e icien , and mo e equi able
heal hca e sys em.
Compliance wi h e hical s anda ds
Disclosu e o con lic o in e es
No con lic o in e es o be disclosed.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2785-2794
2793
Re e ences
[1] Akinwande O. T and Abdullahi (2019). Pe o mance E alua ion o A i icial Immune Sys em Algo i hms o
In usion De ec ion. Jou nal o I e Communica ion o Technology
[2] Ame ican Medical Associa ion. (2020). E hical guidelines o AI in clinical decision-making. Re ie ed om
h ps://www.ama-assn.o g
[3] Ande son, R., & Moo e, T. (2021). Secu i y enginee ing: A guide o building dependable dis ibu ed sys ems (3 d
ed.). Wiley.
[4] Biamon e, J., Wi ek, P., Panco i, N., Reben os , P., Wiebe, N., & Lloyd, S. (2017). Quan um machine lea ning.
Na u e, 549(7671), 195–202. h ps://doi.o g/10.1038/na u e23474
[5] Binns, R. (2018). Fai ness in machine lea ning: Lessons om poli ical philosophy. P oceedings o he 2018
Con e ence on Fai ness, Accoun abili y, and T anspa ency, 149–159.
h ps://doi.o g/10.1145/3178876.3186088
[6] Cen e s o Medica e & Medicaid Se ices. (2021). F aud de ec ion and p e en ion using AI. Re ie ed om
h ps://www.cms.go
[7] Cybe secu i y and In as uc u e Secu i y Agency. (2022). Heal hca e sec o cybe secu i y epo . Re ie ed
om h ps://www.cisa.go
[8] Dwo k, C., & Ro h, A. (2014). The algo i hmic ounda ions o di e en ial p i acy. Founda ions and T ends in
Theo e ical Compu e Science, 9(3–4), 211–407. h ps://doi.o g/10.1561/0400000042
[9] Olu emi, O. D., Ikwuogu, O. F., Kamau, E., Oladejo, A. O., Adewa, A., & Ogun okun, O. (2024). In as uc u e-as-code
o 5g an, co e and sbi deploymen : a comp ehensi e e iew. In e na ional Jou nal o Science and Resea ch
A chi e, 21(3), 144-167. h ps://doi.o g/10.30574/gje a.2024.21.3.0235
[10] Fa hi, E., Golds one, J., & Gu mann, S. (2014). A quan um app oxima e op imiza ion algo i hm. a Xi p ep in
a Xi :1411.4028.
[11] Feldman, M., F iedle , S. A., Moelle , J., Scheidegge , C., & Venka asub amanian, S. (2015). Ce i ying and emo ing
dispa a e impac . P oceedings o he 21s ACM SIGKDD In e na ional Con e ence on Knowledge Disco e y and
Da a Mining, 259–268. h ps://doi.o g/10.1145/2783258.2783311
[12] Fede al T ade Commission. (2021). Using a i icial in elligence and algo i hms. Re ie ed om
h ps://www. c.go
[13] Flo idi, L., Cowls, J., Bel ame i, M., Cha ila, R., Chaze and, P., Dignum, V., ... & Vayena, E. (2018). AI4People—An
e hical amewo k o a good AI socie y: Oppo uni ies, isks, p inciples, and ecommenda ions. Minds and
Machines, 28(4), 689–707. h ps://doi.o g/10.1007/s11023-018-9482-5
[14] Food and D ug Adminis a ion. (2021). A i icial in elligence and machine lea ning in so wa e as a medical
de ice. Re ie ed om h ps://www. da.go
[15] Gisin, N., Ribo dy, G., Ti el, W., & Zbinden, H. (2002). Quan um c yp og aphy. Re iews o Mode n Physics, 74(1),
145–195. h ps://doi.o g/10.1103/Re ModPhys.74.145
[16] Good ellow, I., Bengio, Y., & Cou ille, A. (2016). Deep lea ning. MIT P ess.
[17] Google Heal h. (2021). Fede a ed lea ning o heal hca e aud de ec ion. Re ie ed om h ps://heal h.google
[18] IBM Quan um. (2022). Quan um machine lea ning in heal hca e cybe secu i y. Re ie ed om h ps://quan um-
compu ing.ibm.com
[19] Jessica Beckley (2025). Ad anced Risk Assessmen Techniques: Me ging da a-D i en Analy ics wi h Expe
Insigh s o Na iga e Unce ain Decision-Making P ocesses . In e nal Jou nal o Resea ch Publica ion and Re iews
[20] Bobie-Ansah, D., & A am, H. (2024). Impac o secu e cloud compu ing solu ions on encou aging small and
medium en e p ises o pa icipa e mo e ac i ely in e-comme ce. In e na ional Jou nal o Science & Enginee ing
De elopmen Resea ch, 9(7), 469–483. h p://www.ij i.o g/pape s/IJRTI2407064.pd
[21] Kai ouz, P., McMahan, H. B., A en , B., Belle , A., Bennis, M., Bhagoji, A. N., ... & Zhao, S. (2021). Ad ances and open
p oblems in ede a ed lea ning. Founda ions and T ends in Machine Lea ning, 14(1–2), 1–210.
h ps://doi.o g/10.1561/2200000083