Pancasila International Journal of Applied Social Science E-ISSN 2988-0750 P-ISSN 2988-0769 Volume 3 Issue 03.
September 2025.
Pp.
DOI: https://doi.
org/10.
59653/pancasila.
Copyright by Author Global Strategies in Fraud Prevention and Detection: A Systematic Review Tussi Sulistyowati1*.
Maria Yovita R Pandin2.
Amiartuti Kusmaningtyas3 Universitas 17 Agustus 1945 Surabaya.
Indonesia1 Universitas 17 Agustus 1945 Surabaya.
Indonesia2 Universitas 17 Agustus 1945 Surabaya.
Indonesia3 Corresponding Email: mytussi@gmail.
Received: 01-06-2025 Reviewed: 03-07-2025 Accepted: 15-08-2025 Abstract This study used a qualitative descriptive design of a systematic review of Scopus-indexed journals on fraud detection and prevention from 2021-2025, and the result was 39 relevant The results highlight that fraud prevention and detection both require a harmonized approach through internal auditing, internal controls, technology solutions, and organizational factors like leadership and professional skepticism.
The use of sophisticated technologies such as machine learning, deep learning, and big data analytics significantly enhances detection capabilities, especially in the case of financial transactions.
Besides internal control systems, auditor skills, and whistleblower systems, there are also significant roles to be played.
Ethical concerns, such as privacy and transparency issues within AI-driven systems, have been noticed as well.
Managerial implications consist of keeping organizational internal controls robust, utilizing technological tools, and encouraging a culture of skepticism as well as ethical Future research may focus on the long-term effectiveness of these methods, such as ethical considerations in AI, sectoral applications, and the cost-effectiveness of implementing these solutions in resource-constrained environments.
Keywords: Fraud Prevention.
Fraud Detection.
Systematic Review.
Artificial Intelligence.
Internal Control.
Introduction Fraud is a universal and evolving threat across sectors, causing massive financial and reputational losses globally.
(Karpoff, 2.
Along with growing online transactions, complex financial structures, and internationalized business processes, fraud risk has become more sophisticated and difficult to detect (Pham & Vu, 2.
To respond to it, governments, institutions, and organizations globally have adopted several mechanisms for preventing and detecting fraud (Taherdoost, 2.
These range from traditional internal controls and auditing Pancasila International Journal of Applied Social Science techniques to advanced technology interventions such as artificial intelligence (AI), machine learning (ML), and big data analytics (Qatawneh, 2.
There has been increased academic interest in recent years in identifying the effectiveness, usability, and impact of these measures, particularly in the wake of increasing digitalization and regulatory demands (Amankwah-Amoah et al.
, 2021.
Meilita Rizkynanda et , 2.
This systematic review combines outcomes from Scopus-indexed papers between 2021 and 2025 to map global trends and innovations in fraud detection and prevention.
categorizing methods by country, industry, and methodology, the study aims to provide a general overview of new best practices and success factors.
Research Questions:
RQ1: What are the methods of fraud prevention that have been reported in Scopus-indexed journals between 2021 and 2025? RQ2: What are the most commonly discussed fraud detection methods in Scopus-indexed literature between 2021 and 2025? Literature Review Fraud is an intentional act of deceit for personal or financial gain, often involving the manipulation of information or systems (Acree, 2021.
Singh et al.
, 2.
It includes forms such as financial, procurement, and identity fraud, affecting both the public and private sectors (Ali & Mohd Zaharon, 2024.
Karpoff, 2021.
Modruan et al.
, 2.
Fraud typically exploits control gaps and is driven by opportunity, pressure, and rationalization, as the fraud triangle defines (Kagias et al.
, 2.
Its consequences include financial loss, damage to reputation, legal sanctions, and erosion of public trust (Gottschalk & Hamerton, 2.
As fraud evolves, becoming increasingly complex, organizations must continue to strengthen their detection and prevention measures to protect their assets and uphold integrity (Taherdoost, 2.
Fraud prevention involves the proactive reduction of the risk of fraud before it being It involves effective internal controls, ethical culture, governance, risk assessment, training employees, and reporting mechanisms (Taherdoost, 2.
Emerging technologies like big data.
AI, and e-procurement have raised the level of real-time identification of vulnerabilities (Raghul et al.
, 2.
Fostering integrity, whistleblowing, and leadership support also play significant roles (Zakiy & Satyarini, 2.
Effective prevention not only guards assets but ensures public trust as well, especially in financial and public institutions (Renigier-Bilozor et al.
, 2.
Detection of fraud is identifying suspicious transactions or activity patterns that may indicate wilful deception (Hilal et al.
, 2.
Historically based on internal control, whistleblowing, and physical audits, detection has become smarter with the technology of complex fraud and big data (Putra et al.
, 2.
Detection nowadays employs machine learning, data mining, neural networks, and AI to analyze masses of data, recognize patterns, and adapt when new fraud ruses emerge (Olushola & Mart, 2.
These technologies improve speed and accuracy, making fraud detection essential across industries like banking, insurance, ecommerce, and the public sector (Al-Hashedi & Magalingam, 2.
Besides preventing financial loss, it also improves stakeholder trust and regulatory compliance (MoriN et al.
, 2.
Global Strategies in Fraud Prevention and Detection: A Systematic Review Research Method This study employed a qualitative descriptive design with a systematic literature review following PRISMA guidelines (Anwar et al.
, 2025.
Pardosi et al.
, 2025.
Rethlefsen & Page.
Sulistyowati.
Anwar, et al.
, 2025.
Sulistyowati.
Pardosi, et al.
, 2025.
Sulistyowati & Husda, 2023.
Sulistyowati & Sukati, 2.
of Scopus-indexed journals concerning fraud prevention and detection between 2021 and 2025.
The initial search via the Scopus database was conducted on May 13, 2025, with research article title querying keywords: "Fraud Prevention" OR "Fraud Detection" OR "Anti-Fraud Strategies" OR "Systematic Literature Review" OR "Global Governance" OR "Internal Control Systems" OR "Risk Management" OR "Forensic Auditing" OR "Financial Crime" OR "Technology in Fraud Detection.
" There were 58,536 papers from this search.
The search was also filtered by the following: publication years .
1Ae2.
, subject areas ("Business.
Management and Accounting" and "Economics.
Econometrics and Finance"), document type ("Article"), publication stage ("Final"), keyword (Fraud Detection.
Fraud Prevention"), source type ("Journal"), language ("English"), and open access availability.
The advanced search query was:
TITLE ( "Fraud Detection" OR "Fraud Prevention" OR "Anti-Fraud Strategies" OR "Systematic Literature Review" OR "Global Governance" OR "Internal Control Systems" OR "Risk Management" OR "Forensic Auditing" OR "Financial Crime" OR "Technology in Fraud Detection" ) AND ( LIMIT-TO ( SUBJAREA, "BUSI" ) OR LIMIT-TO ( SUBJAREA, "ECON" ) ) AND ( LIMIT-TO ( DOCTYPE, "ar" ) ) AND ( LIMIT-TO ( PUBSTAGE, "final" ) ) AND ( LIMIT-TO ( SRCTYPE, "j" ) ) AND ( LIMIT-TO ( LANGUAGE, "English" ) ) AND ( LIMIT-TO ( OA, "all" ) ) AND ( LIMIT-TO ( EXACTKEYWORD, "Fraud Detection" ) OR LIMIT-TO ( EXACTKEYWORD, "Fraud Prevention" ) ).
The last search resulted in 39 relevant articles to be included in the subsequent analysis.
The flow diagram is presented Identification Records identified through a title-based database search using keywords such as "Fraud Detection," "Fraud Prevention," and "Anti-Fraud Strategies": n=58,536 Excluded n=58,485 Screening Records identified .
1Ae2.
using filters for subject area, article type, publication stage, journal source, language, open access, and keywords ("Fraud Detection" or "Fraud Prevention"):
n = 39 Excluded Eligibility Included Records identified after full-text screening:
n=39 Studies included in the systematic literature review:
n=39 Figure 1.
PRISMA Flow Diagram of the Literature Review Method Source: (Anwar et al.
, 2025.
Pardosi et al.
, 2025.
Sulistyowati.
Anwar, et al.
, 2025.
Sulistyowati.
Pardosi, et al.
, 2.
Pancasila International Journal of Applied Social Science Result and Discussion The 39 items in Table 1 encompass a wide variety of methodologies, methods, and fraud detection and prevention settings by industry and country.
One general trend in the majority of the studies (Aslam et al.
, 2022.
Baesens et al.
, 2021.
Xu et al.
, 2023.
Yang et al.
is using machine learning (ML), deep learning (DL), or AI techniques to maximize precision and effectiveness in fraud detection, particularly in credit card, loan, medical insurance, and online transactions.
Other research (Khikmah et al.
, 2023.
Riadi et al.
, 2025.
Wahidahwati & Asyik, 2.
also focuses intensely on human factors such as auditor ethics, ability, political will, and leadership in enhancing fraud detection and prevention in public Others (Gabrielli et al.
, 2024.
Mardjono et al.
, 2.
speak of the growing importance of big data and forensic accounting, arguing that technological competence can enhance or enable the efficacy of internal controls.
However, the studies differ widely in topic and scope.
For instance, .
an Bekkum & Borgesius, 2.
and (Gabudeanu et al.
, 2.
have touched on legality and morality more thoroughly in the context of the war of data privacy against surveillance in fraud systems.
Some of them put a special emphasis on systemic and policy-driven analysis (Baumgyrtler et al.
Haliah et al.
, 2.
, and others are algorithm performance-driven or technical.
The regional breakdown also reflects regional interests: Indonesian studies predominantly focus on internal audit, governance, and public accountability, while European and Chinese ones rather apply state-of-the-art AI methods.
Lastly, a few like (Putra et al.
, 2.
and (Joseph et al.
define content analysis or literature reviews rather than empirical testing owing to methodological heterogeneity.
Overall, while a notion appears to exist that technological and institutional measures need to be used to combat fraud, and everyone agrees with this fact, heterogeneity between data, methods, and context goes to helps point to the intricacy of fraud Table 1.
Scopus-Indexed Articles on Fraud Prevention and Detection .
1Ae2.
Author (Yea.
Country Scope Method Data Real (Baesens et , 2.
Belgium
Fraud detection using ML Data
& ML
(Aslam et al.
Pakistan Auto insurance fraud detection ML (SVM.
Logistic.
NB).
Boruta Public auto (Van Belle et , 2.
Belgium
Credit card
fraud detection Network RL
(CATCHM)
Real-life credit card
Finding Improved performance and interpretability using data and feature SVM has the highest
features: fault, policy, age
CATCHM
outperforms others.
reduces manual feature engineering, uses transaction Global Strategies in Fraud Prevention and Detection: A Systematic Review Author (Yea.
an Bekkum & Borgesius.
Country Scope Method Data Netherlands Legal implications of fraud detection Legal case Dutch SyRI court case (Velasco et , 2.
Brazil
Procurement fraud risk
DSS, data
graph theory Public
($50B)
(Wahidahwati & Asyik.
Indonesia Auditor traits & fraud detection Survey & 57 auditors from East Java Real fraud (Xu et al.
China General fraud Deep Boosting Decision Trees (DBDT) (Mqadi et al.
South Africa Credit card fraud with imbalanced data SMOTe classical ML Imbalanced credit card (Chen et al.
China Loan fraud with Hierarchical multi-task Real-world (Putra et al.
Indonesia Fraud strategy review Literature 90 journal .
0Ae2.
(Farbmacher et al.
, 2.
Germany Health insurance fraud Deep Private insurer data (Gabudeanu et , 2.
Romania Transactional fraud vs.
privacy in the Legal 425 survey China Financial fraud Literaturebased Secondary (Miao, 2.
Finding SyRI ruled illegal.
stresses privacy, transparency, and data protection DSS aids identifies fraud patterns .
collusion, conflicts of interes.
Experience, ethics, skepticism, and significantly affect fraud detection DBDT improves accuracy and interpretability by merging neural nets with boosting SMOTe improves the model ability to detect positive .
classes Fraud broken into improves accuracy and handles info Internal audit, risk whistleblowing, big data influence fraud prevention via Deep learning model adapted from text conventional ML in unstructured claim Explores tension between fraud detection and data Traditional models struggle with learning offers promise in feature extraction and pattern recognition.
Pancasila International Journal of Applied Social Science Author (Yea.
Country Scope Method Data Indonesia Forensic accounting.
Big Data, and internal control Quantitative.
SEM
331 auditors in Indonesia Quantitative (SPSS) 65 internal (Mardjono et , 2.
(Khikmah et , 2.
Indonesia Fraud prevention in (Joseph et al.
Malaysia & Indonesia Fraud disclosure on university Content (Mappanyukki et al.
, 2.
Indonesia Emotional and skepticism Quantitative (SmartPLS) 42 SKPD staff in Gowa Regency (Junaidi et al.
Indonesia Political skill.
Big Data, and fraud detection Quantitative (SEM via SmartPLS) 147 auditors (BPKP & BPK) (Detthamrong et al.
, 2.
Thailand Banking fraud ML model Bank (Ghrib et al.
Global Card-based fraud detection Deep (BiLSTM BiGRU) Card (Pramono.
Indonesia
Fraud
Qualitative CFA
30 BPK RI
(Silalahi et al.
Indonesia E-procurement and fraud Quantitative.
SmartPLS officers in Riau (Nguyen et , 2.
Vietnam Internal audit and fraud Quantitative.
SPSS &
SmartPLS 325 joint stock firms Finding BDA mediates the link between COSO/internal control & intent to use forensic IC BDA aids fraud Internal audit and improve fraud Malaysian univ.
Stronger in audit/bursary.
Indonesians are better in governance/policy.
overall, low disclosure levels.
EI affects fraud prevention via skepticism does not moderate the PC-FP Political skill and Big Data positively influence fraud detection ability.
CatBoost outperforms others.
ensemble methods and data sampling enhance accuracy.
Ensemble model detection rate.
traditional ML Auditor experience and task-specific knowledge are key to fraud prevention.
E-procurement and internal control significantly reduce fraud in government Internal audit quality, team independence, and leadership support Global Strategies in Fraud Prevention and Detection: A Systematic Review Author (Yea.
Country Scope Method Data Quantitative.
PLS
61 internal (North Sumatra Inspectorat.
(Lubis et al.
Indonesia Fraud prevention in (Gabrielli et , 2.
Italy Big data in Qualitative.
Interviews 17 forensic (Baumgyrtler et al.
, 2.
EU Structural & Investment Funds Quantitative, 454 EU 4Ae2.
(Kim, 2.
South Korea Online sales fraud detection Deep Smart supply chain Expertbased Financial & opinion data (Istanbul Exchang.
(Benligiray et , 2.
Turkey Fraud detection historical data (Riadi et al.
Indonesia Auditor fraud detection Quantitative.
PLS-SEM
(Zhu et al.
China
Financial fraud
via supply chain
GNN
eHGCN) Multi-year supply chain (Haliah et al.
Indonesia Political will.
IT, and fraud Quantitative.
PLS-SEM
325 village officials in South Sulawesi (Yaseen & AlAmarneh.
UAE & Qatar AI fraud detection in Quantitative.
PLS-SEM &
MGA
409 bank Finding improve audit effectiveness, which enhances fraud Internal audit and internal control significantly prevent audit quality has no effect.
Big data enhances fraud detection via visual analytics and deeper analysis .
High GDP and transparency levels improve fraud
states are more The DL model
detects regions and patterns of sales fraud based on customer/payment New fraud detection scoring performs without needing large training Audit quality mediates the influence of ethics, professionalism, and competence on fraud
IEHGCN
traditional methods.
fraud propagates through supplier Political will improves fraud prevention but not financial reporting IT improves both fraud prevention and reporting quality.
Transparency builds trust, the main driver of AI adoption.
fairness perception reduces bias.
Pancasila International Journal of Applied Social Science Author (Yea.
Country Scope Method Data Finding compliance supports (Cholakov & StoyanovaDoycheva, (Achmad et , 2.
(Damayanti & Agustia.
Bulgaria AI-enhanced fraud detection (FraudDetecto.
Qualitative.
Software ChatGPT Indonesia Forensic skills and fraud Quantitative.
Regression 537 external Indonesia Commitment, religiosity, and Quantitative.
SEM-PLS
(BPK)
Indonesia Whistleblowing, integrity, and fraud prevention Quantitative.
Multiple 70 auditors in South Sulawesi Makassar Indonesia Village financial system and fraud Quantitative.
Regression 51 village officials in Pinrang Quantitative.
MoE
DNNSMOTE
Public credit card dataset Quantitative.
SEM-PLS
117 internal (AIGIA) (Paranoan et , 2.
(Usman & Sundari, (Yang et al.
China Credit card fraud detection using AI (Nadirsyah et , 2.
Indonesia IAF.
IC.
FP,
and governance AI integration (ChatGPT) improves fraud detection architecture supports Communication and auditing skills enhance selfefficacy.
GAS
boosts fraud whistleblowing has no moderating All commitment types and religiosity positively impact responsibility for fraud detection.
Whistleblowing and significantly reduce culture is crucial.
Financial system, transparency, and internal control all significantly prevent Combining MoE and DNN-SMOTE enhances fraud detection, achieving high accuracy and balance (MCC = IAF improves IC and governance.
fraud prevention mediates IAFAe governance link, but IC does not.
Source: Scopus, as of May 13, 2025 RQ1: What are the methods of fraud prevention that have been reported in Scopus-indexed journals between 2021 and 2025? Between 2021 and 2025.
Scopus-indexed journals reported a variety of fraud prevention practices that reflect a growing convergence of technology, governance processes, and Global Strategies in Fraud Prevention and Detection: A Systematic Review institutional reforms.
One such practice is the application of internal audit and internal control Studies in Indonesia (Khikmah et al.
, 2023.
Lubis et al.
, 2024.
Nadirsyah et al.
confirmed that these mechanisms are essential in mitigating fraud risks, particularly in government institutions.
The other interesting trend is the use of digital solutions such as eprocurement systems (Silalahi et al.
, 2.
and village financial information systems (Usman & Sundari, 2.
, which enhance transparency and accountability.
Organizational and individual-level factors are also emphasized.
for example, transformational leadership, auditor capability, professional skepticism, and political will (Haliah et al.
, 2025.
Mappanyukki et al.
have been identified to be linked with more successful fraud prevention.
Whistleblowing mechanism and ethical culture were also established to reduce fraud occurrence (Paranoan et , 2.
Additionally, big data and forensic accounting is also a strategic prevention method (Gabrielli et al.
, 2024.
Mardjono et al.
, 2.
that offers advanced insight into abnormalities.
These findings suggest that fraud prevention requires a multi-dimensional approach combining robust internal processes, digital technology, and human resource development.
RQ2: What are the most commonly discussed fraud detection methods in Scopus-indexed literature between 2021 and 2025? During the period 2021-2025, fraud detection methods in Scopus-indexed papers have been focused primarily on machine learning (ML), deep learning (DL), forensic accounting, internal audit, and big data analytics.
Most studies employed ML and DL algorithms, such as Support Vector Machines (SVM).
Boosting Trees.
BiLSTM-GRU models, and neural networks (Baesens et al.
, 2021.
Ghrib et al.
, 2024.
Xu et al.
, 2023.
Yang et al.
, 2.
These methods enhance accuracy, scalability, and the ability to detect faint fraud patterns, particularly for financial transactions, credit card data, and health insurance claims.
Feature engineering and oversampling approaches like SMOTe (Mqadi et al.
, 2.
have been combined in some studies to tackle imbalanced datasets.
Others introduced new models, such as CATCHM (Van Belle et al.
, 2.
and hierarchical multi-task learning (Chen et al.
, 2.
, that outperform traditional classifiers and reduce human effort.
As a complement to the technological solutions, literature also emphasizes the significance of internal control systems, auditor knowledge, whistleblowing systems, and organizational ethics (Paranoan et al.
, 2024.
Riadi et al.
, 2025.
Wahidahwati & Asyik, 2.
Indonesian studies particularly highlight the internal audit role, auditor skepticism, and leadership style in supporting fraud prevention and detection (Damayanti & Agustia, 2024.
Khikmah et al.
, 2.
Moreover, big data analytics and supply chain knowledge graphs (Mardjono et al.
, 2024.
Zhu et al.
, 2.
reveal new possibilities for the detection of networkbased and intricate fraud.
Legal and ethical concerns were also brought up, specifically on the side of privacy and transparency in AI-driven systems (Gabudeanu et al.
, 2021.
van Bekkum & Borgesius, 2.
, indicating mounting debate about the surveillance-civil rights balance.
Finally, the most universally debated methods incorporate sophisticated analytics (AI/ML/DL), forensic auditing, strong internal controls, and ethical governance models to enhance fraud detection efficiency in numerous business sectors and geographical locations.
Pancasila International Journal of Applied Social Science Conclusion Fraud prevention and detection methods analyzed in Scopus-published journals from 2021 through 2025 reveal that fraud prevention and detection both require a combined, multidimensional approach.
Internal auditing, internal controls, digital solutions like e-procurement and village financial systems, and organizational traits such as leadership and professional skepticism were critical to reduce fraud risk for fraud prevention.
The synergy between forensic accounting and big data also turned out to be a powerful tool in detecting potential threats.
These findings emphasize the need for synergy between good governance practices and technological innovation and human resource capacity building in anti-fraud activities.
As for fraud detection, the main methodologies focus on high-tech methods such as machine learning (ML), deep learning (DL), and big data analytics, which provide higher precision and scalability in the detection of sophisticated fraud patterns, particularly in financial transactions.
Moreover, internal control systems, auditor competencies, and whistleblowing procedures were emphasized as supporting considerations in the delivery of effective fraud detection.
Legal and ethical concerns, particularly privacy and transparency issues within AI-driven systems, were also noted as salient ones.
In general, both fraud prevention and detection methodologies stress the need for ensuring a balance between innovation in technology and efficient regulation and ethical reasoning in combating fraud across industries.
Declaration of conflicting interest The authors declare that there is no conflict of interest in this work.
References