Integrating Artificial Intelligence in organizational cybersecurity: Enhancing consumer data protection in the U.S. Fintech Sector

 Corresponding author: Oluwabiyi Oluwawapelumi Ajakaye ORCID: 0009-0000-4014-4285 Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0. Integrating Artificial Intelligence in organizational cybersecurity: Enhancing consumer data protection in the U.S. Fintech Sector Oluwabiyi Oluwawapelumi Ajakaye 1, *, Ayobami Gabriel Olanrewaju 2, David Fawehinmi 3, Rasheed Afolabi 4 and Gold Mebari Pius-Kiate 5 1 Department of Telecommunications Engineering, University of Sunderland, Tyne and Wear, United Kingdom. 2 Department of Business, Western Governors University, Salt Lake City, Utah, USA. 3 Department of Business, Law and Politics, University of Hull, Kingstom, United Kingdom. 4 Department of Information Systems, Baylor University, Waco, Texas, USA. 5 Department of Computer Science and Information Systems, Pace university, New York, USA. World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 Publication history: Received on 14 March 2025; revised on 20 April 2025; accepted on 22 April 2025 Article DOI: https://doi.org/10.30574/wjarr.2025.26.1.1421 Abstract Financial technology (fintech) companies face escalating cyber threats that jeopardize consumer data. This research investigates how integrating artificial intelligence (AI) into organizational cybersecurity can enhance consumer data protection in the U.S. fintech industry. We pose key questions on AI’s role in threat detection, its current use cases and challenges in fintech cybersecurity, and the effectiveness of deep learning models in preventing data breaches. A comprehensive literature review reveals that AI techniques – particularly deep learning models like Long Short-Term Memory (LSTM) networks and Transformers – are increasingly applied for intrusion detection, fraud mitigation, and threat intelligence in fintech cybersecurity. However, challenges such as adversarial attacks, data bias, regulatory constraints, and implementation costs persist. To address our research questions, we develop an AI-driven cybersecurity methodology applying LSTM and Transformer models to recent U.S. fintech breach datasets and a benchmark intrusion dataset. Real-world breach data from 2018–2023 (e.g., the Verizon VERIS breach database and public disclosures) and a modern intrusion detection dataset are used to train and evaluate the models. The LSTM-based model and Transformer-based model are assessed on their accuracy, detection speed, and impact on breach prevention. Results show that both models achieve high detection rates (over 98–99% accuracy) in identifying malicious activities, with the Transformer slightly outperforming the LSTM in precision and recall. These AI models dramatically reduce incident response times and flag threats that may otherwise go undetected, aligning with industry reports that organizations using security AI contain breaches significantly faster. Discussion of the findings connects these performance gains to improved consumer data protection: earlier and more accurate detection of intrusions allows fintech firms to prevent or mitigate data breaches before sensitive customer information is compromised. We also explore how AI integration must be paired with governance, risk, and compliance (GRC) frameworks to address ethical and regulatory considerations. Conclusion: The study concludes that AI-driven cybersecurity holds great promise for strengthening data protection in fintech by augmenting threat detection capabilities and reducing breach impacts. We provide actionable insights for fintech organizations and researchers, highlighting that while AI can substantially enhance cybersecurity resilience and consumer data safety, a socio-technical approach addressing challenges of trust, transparency, and compliance is essential for successful implementation World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2803 Keywords: Artificial Intelligence (AI); Cybersecurity; Fintech; Consumer Data Protection; Deep Learning Models (LSTM; Transformers); Threat Detection and Prevention 1. Introduction The rapid digitalization of financial services has made fintech companies prime targets for cyberattacks. Fintech firms manage vast amounts of sensitive personal and financial data and offer online services 24/7, which exposes them to a wide array of cyber threats. High-profile data breaches in recent years underscore the severity of this risk. For instance, the 2019 Capital One breach exposed over 100 million customer records due to a cloud misconfiguration, and more recently in 2024, fintech software provider Finastra suffered a breach where attackers exfiltrated 400 GB of data via a compromised file transfer application. Such incidents not only harm consumers through identity theft and fraud, but also erode customer trust and invite regulatory penalties. According to the Identity Theft Resource Center, the financial services sector consistently ranks among the most-breached industries each year Figure 1 illustrates that in 2023 the finance sector experienced 744 reported data compromises – roughly one quarter of all U.S. breaches – second only to healthcare. These statistics highlight a pressing need for more robust cybersecurity measures tailored to fintech’s threat landscape. Figure 1 Data breaches by industry in the U.S. for 2023. The financial sector suffered 744 breaches, reflecting its status as one of the most targeted industries (data from ITRC report) Traditional security controls, while necessary, often struggle to keep pace with modern cyber threats that are increasingly sophisticated and fast-moving. Fintech organizations have begun exploring artificial intelligence (AI) and machine learning as innovative solutions to bolster cybersecurity defenses. AI algorithms can analyze vast streams of network traffic, transaction data, and user behavior logs in real-time, potentially detecting anomalies or attack patterns far more quickly than manual methods. For example, machine learning-driven intrusion detection systems can recognize subtle indicators of attacks amidst enormous data volumes. Likewise, fintech leaders have applied AI to fraud prevention – PayPal’s AI engines, for instance, scan millions of transactions in real-time to block suspicious activity before it results in fraud. Early adoption in the financial sector suggests AI can significantly improve threat detection speed and accuracy, thus enhancing consumer data protection by stopping breaches early. However, integrating AI into an organization’s cybersecurity program is not straightforward. It raises important research questions about efficacy, implementation challenges, and outcomes for data protection. This study focuses on World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2804 U.S. fintech companies and examines how AI techniques (specifically deep learning models) can be leveraged to enhance consumer data security. We formulate the following research questions (RQs) to guide our investigation: • RQ1: What are the current use cases of AI in fintech cybersecurity, and how do these applications contribute to protecting consumer data? • RQ2: What challenges and limitations do fintech organizations face in implementing AI-driven cybersecurity solutions (e.g., technical, operational, ethical, and regulatory challenges)? • RQ3: How effectively can deep learning models such as LSTMs and Transformers detect or predict cybersecurity incidents in fintech, and in what ways could their deployment improve consumer data protection outcomes compared to traditional methods? To answer these questions, we combine a literature-driven analysis with an empirical evaluation of AI models on relevant cybersecurity datasets. The literature review (Section 2) synthesizes findings from peer-reviewed studies on AI applications in cybersecurity – intrusion detection, fraud detection, threat intelligence – with particular attention to the fintech context. We also identify the prevailing challenges that might hinder AI integration (e.g., adversarial attacks on AI, data privacy concerns, need for model transparency). Next, the methodology (Section 3) details our approach in applying deep learning models to recent cybersecurity incident data. We describe the data sources, model architectures, and evaluation metrics used to assess AI performance in detecting breaches or anomalies. In Section 4, we present results from our experiments, including quantitative performance of the AI models and qualitative observations on how AI could mitigate real-world breach scenarios. Section 5 provides an in-depth discussion, interpreting how our findings answer the RQs and offering implications for fintech industry practice (such as the expected reduction in breach detection time and improvements in defensive capabilities with AI). We also discuss how organizations can address the challenges identified, for example by combining AI with governance and risk management strategies. Finally, Section 6 concludes the paper by summarizing the key insights: we find that integrating AI into fintech cybersecurity can substantially enhance consumer data protection by augmenting threat detection and incident response, but it requires careful implementation to overcome limitations. We also suggest directions for future research, including the development of explainable and adversary-resistant AI for cybersecurity. Overall, this work provides a comprehensive examination of AI’s role in fintech cybersecurity. It offers evidence that advanced AI techniques, when properly harnessed, can strengthen organizational defenses and better safeguard sensitive customer data in the digital finance era – a contribution that is increasingly critical as cyber threats continue to grow in frequency and complexity. 2. Literature Review 2.1. Cybersecurity Threat Landscape in Fintech Fintech companies operate at the intersection of finance and technology, making them uniquely vulnerable to a broad spectrum of cyber threats. Their reliance on digital platforms, APIs, and cloud services, combined with handling of valuable financial data, attracts both financially motivated cybercriminals and other threat actors. A recent systematic review by Javaheri et al. (2024) identified 11 central cyber threats facing fintech firms, highlighting the prevalence of data breaches, malware attacks, phishing, ransomware, and insider threats. Common attack vectors include social engineering (tricking employees or customers into divulging credentials), exploitation of unpatched software vulnerabilities, and abuse of weak authentication or access controls. Phishing is particularly rampant in the fintech space – attackers impersonate banks or payment services to steal login credentials or trick users into installing malware . These techniques can lead to unauthorized access to systems containing personal data or account information. Fintech firms also increasingly face ransomware attacks, where malware encrypts critical data and demands payment; such incidents have surged in recent years and can cripple operations if backups are inadequate. For example, a successful ransomware attack on a fintech’s infrastructure might force the business offline and potentially expose or destroy customer records, causing both financial and reputational damage. Another important threat category is direct data breaches through hacking. Fintech companies store large quantities of personally identifiable information (PII), banking details, and transaction records, making them “prime targets for hackers”. Breaches can occur via external attacks (e.g., exploiting a web application vulnerability as in the Equifax 2017 breach) or via insider misuse. The consequences of such breaches are severe: besides immediate financial losses and customer harm, firms may face regulatory fines and legal liabilities under data protection laws. In the U.S., regulations like the Gramm-Leach-Bliley Act (GLBA) and state data breach notification laws impose strict duties to safeguard customer data and disclose breaches, so a fintech breach often triggers expensive remediation and compliance costs. Indeed, studies show data breaches can have long-term impacts on customer trust – fintech users are likely to switch providers if they feel their data is not secure, and loss of trust can significantly erode a company’s market value. World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2805 Compounding the issue, threat actors are continually evolving their tactics. Emerging threats include the use of AI by attackers themselves. Cybercriminals have begun to leverage AI and automation to launch more sophisticated attacks, such as AI-driven phishing that crafts highly personalized bait, or malware that can adapt its behavior to evade detection in real-time. For example, malware augmented with AI might dynamically modify its signatures to avoid antivirus software, or criminals might use machine learning to find and exploit new vulnerabilities faster. Such AI-powered attacks pose a new challenge for defenders, as they can defeat static security measures and require more adaptive defenses. Fintech is also exposed to threats via the supply chain and third-party dependencies. Many fintech services rely on third-party software (e.g., open source libraries, cloud hosting) and integrations with banking partners. A vulnerability or breach at a third-party (such as the 2023 MOVEit file-transfer software zero-day that led to numerous downstream data breaches) can indirectly compromise fintech data. In 2023, the Kroll Data Breach Outlook noted that third-party risk was a leading cause of incidents, and the finance sector was heavily impacted by a supply-chain attack wherein a ransomware gang exploited a common file-transfer tool, affecting multiple financial institutions. This interconnectedness means fintech firms must account for not only their own security, but also the security of vendors and partners. In summary, the threat landscape for fintech is both broad and dynamic. Fintech organizations face the full gamut of cyber-attacks seen in traditional finance (fraud, account takeovers, data theft) as well as technology-sector attacks (DDoS, exploits of software flaws). Table 1 provides a brief overview of representative cybersecurity incidents in fintech over the last five years to illustrate the range of threats and impacts. Table 1 Major Fintech-Related Data Breaches (2018–2023) Incident (Year) Company/Service Records Affected Cause of Breach Capital One breach (2019) Capital One (bank/fintech) ~106 million customers Misconfigured AWS cloud storage exploited by hacker (server-side request forgery) First American Financial breach (2019) First American (fintech RE) 885 million records Web application logic flaw exposing documents without authentication Dave.com breach (2020) Dave (digital bank app) ~7.5 million users Hacking of third-party service provider, leading to credential compromise Robinhood breach (2021) Robinhood (stock trading) ~7 million customers Social engineering of customer support, allowing attacker to obtain user data Cash App Investing incident (2022) Block (Cash App) ~8 million customers Insider threat – former employee downloaded reports containing user stock data Finastra breach (2024) Finastra (fintech software) Unknown (65 notified in MA) Compromised file transfer application by attacker; 400 GB of data stolen (no ransomware) These examples underscore recurring patterns: configuration errors or unpatched software leading to breaches (Capital One, First American), third-party or insider risks (Dave, Cash App, Finastra), and social engineering that bypasses technical controls (Robinhood). The high frequency and impact of such incidents have pushed the fintech industry and regulators to seek stronger defenses – which is where AI has emerged as a promising tool, as discussed next. 2.2. Applications of AI in Fintech Cybersecurity Artificial intelligence has rapidly become a cornerstone of next-generation cybersecurity solutions. In particular, machine learning (ML) and deep learning techniques enable security systems to analyze complex data and detect threats with a speed and sophistication that complements human analysts. In the fintech domain, organizations are applying AI across several key cybersecurity use cases: • Intrusion and Anomaly Detection: Detecting network intrusions and system anomalies in real time is crucial for preventing data breaches. AI-driven Intrusion Detection Systems (IDS) use ML algorithms to model normal versus malicious behavior. For example, deep learning models can learn patterns of legitimate network traffic and flag deviations that might indicate a breach attempt. LSTM neural networks, which excel at sequence learning, have been used to identify suspicious sequences of events in network logs or user activity. Prior research has demonstrated that optimized LSTM models can achieve high accuracy in detecting intrusions World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2806 while reducing false alarms. In one study, a tuned LSTM-based IDS achieved over 97% detection accuracy on benchmark datasets, outperforming earlier machine-learning IDS approaches that often suffered from high false-positive rates. Similarly, Transformer-based models (which rely on self-attention mechanisms) have been adapted for cybersecurity to great effect. A recent work by Ataa et al. (2024) developed a Transformer encoder model for network intrusion detection in cloud environments and reported ~99% classification accuracy, slightly higher than a convolutional LSTM hybrid model on the same data. The ability of Transformers to capture long-range dependencies in data is beneficial for modeling complex attack patterns. These advancements indicate that AI can markedly improve the detection of unauthorized access or abnormal activities in fintech systems, enabling security teams to respond to incidents before they escalate into fullblown breaches. • Fraud Detection and Transaction Monitoring: Payment fraud and identity theft are major concerns in fintech (e.g., credit card fraud, fraudulent account openings, and money laundering). AI has become indispensable in detecting fraud in real-time by analyzing transaction data. Machine learning models (such as decision trees, random forests, gradient boosting) have long been used to flag potentially fraudulent transactions based on rules and patterns. More recently, fintech companies have deployed deep learning for fraud detection – for instance, using neural networks to evaluate dozens of features of each transaction (amount, location, device, past behavior, etc.) to predict fraud probability. PayPal’s deployment of AI is a prominent example: using a combination of neural networks and anomaly detection algorithms, PayPal reported it can identify fraudulent transactions within milliseconds and block them before completion, reducing fraud loss rates significantly. Large banks and card networks similarly use AI to score transactions; the model’s ability to learn subtle, nonlinear correlations means it catches fraud that simpler rules might miss (for example, complex account takeover schemes involving multiple steps). AI-based fraud detection not only protects consumer assets but also indirectly protects personal data by detecting when an unauthorized user may be leveraging stolen credentials. • User & Entity Behavior Analytics (UEBA): Fintech firms are adopting AI to establish baselines of normal behavior for users, devices, and applications, and then detect anomalies that could signal an insider threat or account compromise. For instance, an AI system might learn that a particular customer typically logs in from Texas and makes small transfers; if suddenly their account initiates a large transfer from overseas, the system will flag it. These behavior-based systems often use unsupervised learning or clustering to discern patterns without needing explicit attack signatures. Many advanced threat prevention systems incorporate UEBA modules powered by AI, which is especially useful for detecting insider threats or subtle breaches where an attacker impersonates a legitimate user. • Threat Intelligence and Phishing Detection (NLP): Another crucial application of AI is in processing unstructured cybersecurity data – an area where Natural Language Processing (NLP) techniques are valuable. AI can automatically ingest threat intelligence reports, news, and forum discussions to identify emerging threats relevant to fintech (such as new malware targeting banking apps). Moreover, NLP is being used to detect phishing and social engineering attempts by analyzing email or message content. For example, ML models can scan incoming emails to employees or customers for phishing indicators (suspicious language patterns, anomalous sender, malicious links). Large language models or transformers fine-tuned for phishing email classification have shown high success in filtering out phish with minimal false positives. This is increasingly important as phishing remains a top initial vector in data breaches. By catching phishing emails or messages before a user falls victim, AI can prevent credential theft that often leads to deeper network infiltration. In addition, AI text analysis helps in fraud prevention (catching scam communications) and compliance (monitoring communications for policy violations). • Security Information and Event Management (SIEM) & Orchestration: AI is also enhancing how security operations centers aggregate and respond to alerts. Traditional SIEM platforms generate large volumes of alerts from logs and events, which can overwhelm analysts. Modern SIEM and security orchestration tools integrate AI algorithms to correlate events and prioritize alerts. For instance, if multiple low-level events (unusual login, followed by a file access, followed by a database query) occur that on their own might not trigger alarms, an AI system can recognize the pattern as potentially malicious when taken together. AI-based correlation and incident scoring reduce noise and highlight the most important threats, making incident response more efficient. Some fintech companies are even exploring automated incident response, where AI not only detects but also initiates containment (such as locking a compromised account or isolating a server) according to predefined playbooks. Overall, current literature and industry reports indicate that AI’s adaptability and learning capabilities make it wellsuited to address the evolving threat landscape in fintech. Intrusion detection using AI has emerged as the most prominent focus area in research, comprising ~13% of publications on AI in cybersecurity, which reflects its critical importance. Other major areas where AI contributes include malware detection, fraud prevention, and threat prediction World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2807 . In fintech organizations, these translate to practical deployments like automated monitoring of network traffic for breaches, real-time fraud scoring in payment systems, and AI-driven security analytics that augment human decisionmaking. Importantly, AI is not viewed as a replacement for human security teams, but as a force multiplier – it can handle the “heavy lifting” of data analysis, surface insights, and even act autonomously for known patterns, allowing security professionals to focus on strategy and on novel or complex threats. Despite these benefits, integrating AI into cybersecurity is not without challenges. We next discuss the limitations and obstacles that fintech firms must consider when deploying AI-driven security systems. 2.3. Challenges and Limitations of AI in Cybersecurity for Fintech While AI offers powerful capabilities, there are several challenges and limitations associated with its use in cybersecurity, particularly in an industry as sensitive as financial services: • Adversarial Attacks on AI: Just as AI can be used to detect attacks, attackers can target the AI models themselves. Adversarial machine learning is an emerging concern; threat actors may attempt to deceive an AI system through specially crafted inputs. For example, an attacker might slightly perturb network traffic patterns or malware code to evade an AI-based detector (often called adversarial evasion) or inject poisoned data during the training phase to corrupt the model’s learning. Fintech security AI systems could be tricked into misclassifying malicious activity as benign, opening a blind spot for attackers. This cat-and-mouse dynamic means that AI models need to be hardened and continuously evaluated against adversarial tactics. Research in explainable AI (XAI) and robust AI is attempting to address this by making models more interpretable and resistant to manipulation. • False Positives and Model Accuracy: Achieving high detection rates without overwhelming analysts with false positives is a delicate balance. Finance is a domain where false alarms carry a cost – for instance, falsely accusing legitimate transactions or user actions can inconvenience customers or interrupt business processes. Earlier machine learning systems often had high false positive rates in IDS applications. Deep learning has improved accuracy, but models still must be fine-tuned for the specific environment to avoid alert fatigue. There is a risk that if an AI system is too sensitive, security teams might start ignoring its alerts (the “boy who cried wolf” effect). Thus, maintaining model precision (high true positive vs false positive ratio) is critical for practical deployment. Our literature review found that optimized models (e.g., using hyperparameter tuning, ensemble methods, etc.) can reduce false positives significantly, but rigorous testing on real fintech data is needed to validate performance before full deployment. • Data Availability and Quality: AI effectiveness depends heavily on data. Fintech companies may have an abundance of certain data (e.g., transaction records) but relatively few examples of actual attacks or fraud cases to learn from. This class imbalance problem can make it hard for supervised models to learn to detect the rare events (the attacks) amidst huge volumes of normal events. If AI models are trained on limited historical incident data, they might not generalize well to new types of attacks. Additionally, acquiring high-quality cybersecurity datasets for training is challenging due to privacy and sensitivity – firms may be reluctant to share breach data, and simulations might not capture all real-world nuances. The VERIS Community Database (VCDB) is one effort to compile thousands of publicly disclosed breach incidents for research, which we leverage in this study. However, as noted in the VCDB documentation, the data can be biased (e.g., overrepresentation of certain sectors like healthcare due to regulatory reporting requirements). Fintech-specific incident data may be underrepresented. To mitigate this, organizations need strategies like data augmentation, transfer learning (using models pre-trained on similar cybersecurity tasks), or federated learning (collaborative model training without sharing raw data) to improve AI models. In practice, large financial institutions have started sharing anonymized fraud indicators among consortiums to bolster AI models – but smaller fintech startups might lack access to such rich data, creating an adoption gap. • Regulatory and Ethical Constraints: In a regulated domain, the use of AI must comply with regulations and uphold customer rights. Regulations like the EU’s GDPR and emerging U.S. state privacy laws require that personal data usage be transparent and fair. If an AI system processes customer data (even for security), it must be secured and its outputs auditable. There is also regulatory attention on the fairness of AI decisions – for example, if an AI flags certain transactions or accounts as high risk, firms need to ensure this does not result in unfair bias against any group of customers. Financial regulators have begun scrutinizing the use of AI (e.g., the U.S. SEC’s guidance on automated digital advice and FRB/FDIC/OCC statements on AI in banking). In cybersecurity specifically, the concern is more on ensuring AI doesn’t violate privacy (for instance, monitoring employee communications with AI could raise workplace privacy issues) and that incident response using AI follows required breach reporting procedures. Additionally, new regulations are being proposed that might mandate explainability for AI decisions in security – i.e., firms may need to explain to an auditor why an AI World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2808 system took a certain action. Black-box models like deep neural networks are notoriously hard to interpret, which poses a challenge. As Achuthan et al. (2024) discuss, ethical concerns and trust in AI remain significant issues; more research is needed in explainable and trustworthy AI for cybersecurity. Fintech companies will likely need to implement AI in a way that humans can oversee and override when necessary, maintaining the “human in the loop” for critical security decisions. • Integration and Operational Challenges: Deploying AI for cybersecurity is not just a technical exercise; it also involves organizational readiness. Smaller fintech startups may lack in-house expertise in data science or cybersecurity AI, making it difficult to build or manage such systems. Purchasing off-the-shelf AI security products is an option, but those might not be tailored to the specific environment. There is also the issue of integrating AI tools with existing security workflows. AI might output a risk score or alert – the security team needs playbooks to act on these alerts. If the integration is poor, the AI tool could be underutilized. Additionally, AI models require continuous maintenance: retraining with new data, updating for new threats, and patching the models themselves. This can strain resources, as noted by industry practitioners interviewed in a U.S. Treasury report on AI in financial sector cybersecurity – many firms proceed cautiously with AI adoption, running pilot projects and ensuring cross-team collaboration (IT, security, compliance, etc.). • False Sense of Security: Finally, an often intangible but important risk is that deploying AI could give organizations a false sense of security if not accompanied by broader cybersecurity improvements. AI is a tool, not a panacea. If a fintech firm relies too heavily on AI and neglects basic security hygiene (patch management, access control, encryption, etc.), it may actually worsen their security posture. Effective cyber defense still requires layered controls (“defense in depth”) where AI is one layer. Misconfigurations or process failures can still cause breaches (for example, an AI system might detect an anomaly but if the incident response process is broken, the breach may not be contained in time). Therefore, implementing AI must be part of a holistic cybersecurity strategy. In summary, while AI brings powerful capabilities to fintech cybersecurity, these limitations mean that organizations must adopt AI thoughtfully. They should combine AI with strong governance, risk management, and human expertise to ensure it truly enhances security. Table 2 summarizes some key challenges of using AI in fintech cybersecurity and approaches to address them. Table 2 Key Challenges in AI-Driven Fintech Cybersecurity and Mitigation Strategies Challenge Description Mitigation Strategies Adversarial ML Attackers crafting inputs to evade or poison AI models – Adversarial training of models with perturbed examples – Model introspection and use of robust architectures – Monitor for model drift or anomalies in model decisions False Positives vs. Misses Tuning sensitivity to minimize false alarms and missed attacks – Threshold tuning and feedback loop with analysts – Ensemble models combining AI with rule-based checks for validation – Use of confidence scoring and only automating high-confidence alerts Data scarcity & imbalance Limited attack data to train models; imbalanced datasets – Use of simulated data and augmentation to supplement real data – Transfer learning from related domains (e.g., using models pre-trained on large cybersecurity data) – Federated learning across institutions to build shared models without sharing raw data Black-box model transparency Difficulty in explaining AI decisions for audits or trust – Implement explainable AI tools (e.g., SHAP values, LIME) to interpret model outputs – Prefer simpler models where appropriate or augment black-boxes with interpretable rules – Maintain human oversight on AI decisions (human-AI teaming) Regulatory compliance Ensuring AI use complies with data protection and financial regulations – Consult compliance teams early in AI deployment design – Anonymize or encrypt sensitive data used in model training (preserve privacy) – Document AI decision processes and validate no discriminatory impact World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2809 Integration & maintenance Operationalizing AI in the security workflow and updating it – Pilot AI systems in parallel with existing monitoring to calibrate – Train security personnel on interpreting AI outputs – Set up regular model retraining schedule and incident post-mortems to refine AI (continuous improvement) Overreliance on AI Neglecting other controls due to confidence in AI – Use AI as one layer in a multi-layer defense strategy – Continue investments in fundamental security controls (network segmentation, least privilege, etc.) – Periodic red-team exercises to test both AI and non-AI controls in unison The literature emphasizes that addressing these challenges is feasible. For instance, case studies have shown that organizations combining AI with strong governance have managed to reduce breach incidents while maintaining compliance. A holistic approach is needed: as suggested by Oluokun et al. (2024), integrating AI with Governance, Risk, and Compliance (GRC) frameworks ensures that AI-driven tools are aligned with regulatory requirements and internal policies. Governance measures, such as clear policies on AI use and defined roles and responsibilities, help in accountable deployment of AI. If done well, the synergy of AI technology with human expertise and governance can yield a robust cyber defense posture for fintech firms. In light of these insights from the literature, our research methodology is designed to explore the practical application of AI models in fintech cybersecurity while cognizant of these challenges. We proceed to describe the methodology, including data collection and model implementation, used to evaluate how AI (specifically deep learning models) can detect security incidents and ultimately protect consumer data in a fintech context. 3. Methodology To investigate the effectiveness of AI in enhancing fintech cybersecurity, we designed a methodology with two main components: (1) Data Collection from real-world cybersecurity incidents and simulated attack data relevant to fintech, and (2) AI Model Application using deep learning techniques (LSTM and Transformer models) to analyze this data for threat detection. Our approach aims to mirror a realistic organizational deployment of AI for cybersecurity, while addressing our research questions about model performance and data protection improvements. 3.1. Data Sources We leveraged two types of datasets to capture both high-level breach incident trends and low-level attack patterns: • Fintech Cybersecurity Incident Dataset (2018–2023): We compiled a dataset of publicly reported cybersecurity incidents affecting U.S. financial institutions and fintech companies over the last five years. This was drawn from the Verizon VERIS Community Database (VCDB) and supplemental public breach disclosures. The VCDB provides structured records of thousands of incidents, including attributes like threat actions, affected assets, and impact. We filtered VCDB records for the Finance and Insurance sector and for incidents from 2018 onward, yielding 650+ incidents. This included notable fintech-related breaches summarized in Table 1 (Capital One, Robinhood, etc.) and many lesser-known cases (e.g., breaches at regional banks, payment processors, credit unions). For each incident, we extracted features such as incident year, attack vector (hacking, malware, misuse, etc.), data types compromised (personal data, financial data, credentials), and whether the incident was caused by internal or external actors. We also labeled incidents by their severity (e.g., number of records compromised) to facilitate supervised learning experiments. This incident dataset allows us to analyze macro-level patterns and also to attempt predictive modeling (for instance, predicting which factors lead to large breaches). All data was sourced from public reports or databases; sensitive details were anonymized or aggregated to respect privacy. • Intrusion Detection Dataset (network attack traces): To evaluate deep learning models in detecting technical attack patterns, we used the CSE-CIC-IDS2018 intrusion detection dataset (an advanced benchmark dataset for network security research). This dataset contains traffic captures and extracted features from a test network environment that includes both benign activity and a variety of attack scenarios (brute-force logins, denial-of-service, web attacks, infiltration, botnet, etc.). The CIC-IDS2018 dataset is curated by the Canadian Institute for Cybersecurity and is widely used for evaluating IDS models; it closely resembles real-world traffic with labeled flows and timestamps. We specifically chose this dataset because it incorporates modern attack vectors that a fintech company might face (including attacks on web services and infrastructure). It provides detailed labeled data at the packet/flow level. From this dataset, we derived a training set and test set of World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2810 network flow records with features such as packet rates, protocol, byte counts, and flags, labeled as either normal or one of multiple attack types. This allows us to train and test our LSTM and Transformer models on a supervised classification task: distinguishing malicious traffic from normal. Although this dataset is not fintechspecific, it offers a controlled benchmark to quantify model detection performance (accuracy, precision, recall) for a broad range of cyber attacks. We consider this a proxy for how the models would perform on fintech network data, under the assumption that fundamental attack patterns (port scans, SQL injection attempts, etc.) are similar across domains. The combination of these two data sources provides a comprehensive view: the incident dataset reflects organizational-level outcomes (breaches, compromised records) and contextual factors, whereas the intrusion dataset provides low-level telemetry for algorithmic detection tasks. By using both, we can examine RQ3 from two angles – can AI predict or classify breach incidents based on organizational data, and can AI detect attacks from technical data streams. Before model training, we performed standard preprocessing on both datasets. The incident dataset, being categorical, was one-hot encoded for features like attack vector and data type, and numeric scales (like company size, records lost) were normalized. We also time-sequenced the incidents by quarter to enable a time-series analysis (for anomaly detection on breach frequencies). The intrusion dataset was cleaned to remove duplicate flows and impute any missing values; continuous features were scaled (using min-max normalization) and categorical protocol flags were encoded. 3.2. AI Model Architectures and Training We implemented two deep learning model architectures aligned with our focus on LSTM and Transformer techniques: 3.2.1. Long Short-Term Memory (LSTM) Model Our LSTM model is designed as an anomaly detection network for sequential data. We constructed it as a multi-layer LSTM autoencoder, which learns to reconstruct normal sequence patterns and flags deviations. This architecture was chosen to detect anomalies in either time-series breach data or sequences of network events. For the incident dataset, we used the LSTM in a time-series forecasting context: the model was trained on the sequence of quarterly incident counts (and features such as average records breached per quarter) to predict future values, with the idea that significant deviation between predicted and actual incidents could indicate an anomaly (e.g., an unusual spike in breaches). For the intrusion data, we applied the LSTM in a classification setting: treating the flow of packets in a session as a sequence, the LSTM processes the sequence and outputs a classification (normal or specific attack type). The model architecture consisted of an input layer matching the feature vector length, two LSTM layers (with 128 and 64 units respectively) to capture temporal patterns, followed by a dense output layer. We trained the classification LSTM model using labeled data (supervised) with a categorical cross-entropy loss, using the Adam optimizer. For the anomalydetection variant (unsupervised), we trained the autoencoder to minimize reconstruction error on a corpus of known “normal” sequences, and set a threshold on reconstruction error to flag anomalies. The training was performed for 50 epochs on the intrusion dataset and 100 epochs on the incident time-series (which was small). We tuned hyperparameters such as learning rate and LSTM layer sizes using a grid search on a validation set. To mitigate overfitting, regularization techniques like dropout (20% dropout between LSTM layers) and early stopping (monitoring validation loss) were applied. The LSTM’s ability to remember long-term dependencies in sequences makes it suitable for capturing the evolving context in a series of events (e.g., a slow ongoing breach or multi-stage attack). 3.2.2. Transformer Model Our Transformer model is designed for high-accuracy supervised classification of security events. We implemented a Transformer encoder architecture similar to those used in recent IDS research. For the intrusion detection task, the Transformer takes as input a sequence of network flow features (we treat each flow or a short window of flows as a sequence tokenized by time) and outputs class probabilities for attack vs normal. The model consists of an embedding layer (to project input features into a learned vector space), followed by multiple self-attention encoder blocks. We used 4 encoder layers, 8 attention heads, and an embedding dimension of 64 in our configuration – these values were chosen based on prior studies that achieved strong results on similar data. The self-attention mechanism of the Transformer allows the model to weigh the relevance of different parts of the input sequence, which is useful to identify, for example, which combination of network features at various time steps signal an attack. After the encoder layers, a global average pooling was applied, then a feed-forward network and softmax layer to produce output probabilities. We trained the Transformer on the CIC-IDS2018 data (supervised multi-class classification) using the Adam optimizer with learning rate 1e-4, and categorical cross-entropy loss. Training ran for 20 epochs (the larger model converged faster than LSTM per epoch due to parallelism of attention). For regularization, we employed dropout in the feed-forward layers and L2 World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2817 • RQ3 (Model Effectiveness and Consumer Data Protection): Our empirical results demonstrated that deep learning models (LSTM and Transformer) are highly effective in detecting cybersecurity incidents relevant to fintech. By achieving over 99% detection rates on simulated attack data, these models vastly outperform legacy detection methods, suggesting a new level of security capability is available. The linkage to improved consumer data protection is clear: by catching intrusions and fraud quickly and accurately, AI helps prevent attackers from accessing sensitive customer information. We discussed how, in concrete terms, this leads to shorter breach durations and less data stolen. One striking statistic from our findings is the potential reduction in breach lifecycle – identifying breaches days or months faster. For consumers, this could be the difference between having their data snatched by criminals versus not at all. Additionally, even when breaches occur, AI can limit their scope, meaning fewer individuals affected. Another aspect is that AI can help comply with data protection principles like data minimization – if AI prevents unauthorized data access, it ensures that personal data isn’t being unnecessarily copied or moved around by malicious actors, thus keeping data only where it should be. Our study also indicates that these benefits are not just theoretical. There are real-world parallels: for instance, Mastercard reported its AI-based fraud systems have reduced false declines and saved millions in fraud losses – aligning with our findings that AI can both tighten security and reduce friction (false positives) for legitimate users. In our context, reducing false positives means genuine user transactions or activities won’t be wrongly blocked as often, providing a smoother user experience while still protecting data. This is an important point: effective security need not come at the expense of user convenience if AI is used wisely. Historically, adding security (like multi-factor authentication, transaction verification steps) added some friction, but AI works in the background, transparently increasing security without bothering the user unless truly necessary. It is worth noting the scope of our experiments: we focused on detection of incidents. Prevention is another area (like using AI to scan code for vulnerabilities or misconfigurations). We didn’t directly test that, but literature suggests AI can aid in those preventive measures too (e.g., code analysis tools with ML). Fintechs could use such tools to catch security flaws in software before deployment, indirectly protecting data by reducing breach opportunities. That extends the protective net of AI beyond real-time monitoring to the whole lifecycle of systems. • Integration with GRC: A theme that emerged is that combining AI with governance and risk management amplifies the benefits. Our discussion of challenges touched on this, but to elaborate: the best outcomes were seen when AI is part of a broader risk strategy. For example, if a fintech adopts an AI system for threat detection, and simultaneously implements a robust incident response plan (as recommended in many cybersecurity frameworks), the synergy means that when AI flags something, the organization is prepared to act quickly. If either element is missing (great detection but poor response, or vice versa), data might still be lost. We cited Oluokun et al. (2024) on integrating AI with GRC– our findings strongly support that recommendation. We would advise fintech firms to not treat AI as a plug-and-play appliance, but to update their policies, training, and drills around it. This includes addressing the ethical aspects: for example, deciding under what conditions AI can autonomously shut down services to contain an attack (affecting availability, which must be weighed against security). Those decisions should be codified in governance documents. • Limitations and Future Research: While our study demonstrates clear benefits of AI, it also has limitations which point to future research directions. One limitation is that our evaluation of model performance was largely on benchmark data; real production environments have more noise and variety. Deploying these models in a live fintech environment might reveal new challenges (for instance, concept drift – the statistical properties of input data may change as user behavior or attacker tactics evolve over time, requiring model updates). Future work could involve longitudinal studies of AI model performance in production, observing how they degrade or improve and how often they need retraining. Another area for future research is • explainable AI in fintech security: developing methods to interpret a Transformer’s alert (e.g., highlight which features or sequence elements led it to tag an event as malicious) so that analysts and auditors can trust and verify the system. This will be increasingly important as regulators may scrutinize AI decisions in finance. Work on integrating AI outputs with existing Security Orchestration, Automation, and Response (SOAR) systems could further bridge the gap from detection to automated response – an area ripe for development (e.g., using AI to not just say “there is an attack” but also suggest or initiate the best containment action). Another consideration is privacy-preserving AI. Fintechs have to be mindful of customer privacy even as they monitor systems. Techniques like federated learning or secure multi-party computation could allow multiple institutions to collaboratively train shared threat models without exposing raw data to each other. This could greatly enhance AI effectiveness (more data to learn from) while respecting confidentiality. Our work didn’t explore that, but it’s a World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2818 promising area especially for industry-wide initiatives against fraud or money laundering where patterns span institutions. • Policy and Regulatory Implications: Given the findings, regulators might encourage responsible use of AI in cybersecurity. Already, the U.S. Treasury has noted AI’s strategic value in fraud detection and risk management in finance. Regulators could incorporate guidance in frameworks like the FFIEC IT Examination Handbook or NIST guidelines specific to the financial sector, recommending that banks and fintechs leverage AI/automation to meet certain security baseline requirements (with the caveat of ensuring human oversight and explainability). If most financial breaches are occurring due to late detection or human error, mandating or strongly incentivizing the use of automated detection could reduce overall consumer harm. Of course, regulators will also keep an eye on ensuring AI doesn’t introduce uncontrolled risks (like algorithmic bias even in security contexts, though less of an issue here than in lending or hiring). Our research provides evidence that, used correctly, AI is a net positive for consumer protection. • Strategic Impact on Fintech Sector: Widespread adoption of AI-driven cybersecurity could become a competitive advantage and an expectation. Fintech companies that invest in these capabilities may gain trust from customers and partners (e.g., showing low incident rates, fast containment in their track record). Conversely, those that lag might suffer more breaches and reputational hits. In time, consumers might even inquire or be informed about the security measures protecting their data – similar to how some tech companies publicize that they use advanced encryption and anomaly detection to safeguard user information. AI in cybersecurity might thus become part of the value proposition of fintech products (implicitly or explicitly). In conclusion, our study finds that integrating AI into organizational cybersecurity significantly strengthens the protection of consumer data in fintech. It confirms many theoretical benefits with practical evidence: improved detection accuracy, speed, and breadth. It also clarifies the path to implementation, highlighting the need to manage challenges like adversarial robustness and compliance. Fintech organizations stand to greatly benefit from these technologies, and ultimately, so do their customers whose data and financial assets will be more secure. With thoughtful governance, continuous improvement, and adherence to ethical standards, AI-driven cybersecurity can usher in a new era of resilience in the financial technology sector. 6. Conclusion The convergence of finance and technology in fintech has brought tremendous convenience and innovation to consumers, but it has equally attracted sophisticated cyber threats that put sensitive personal and financial data at risk. This research has provided a comprehensive examination of how artificial intelligence – particularly deep learning models like LSTM networks and Transformers – can be integrated into organizational cybersecurity to enhance the protection of consumer data in the U.S. fintech sector. Our work addressed key research questions by combining an extensive literature review with empirical modeling on real-world inspired data. We found that AI techniques are already proving their value in fintech cybersecurity: they are used for real-time intrusion detection, fraud prevention, user behavior analytics, and threat intelligence, all of which contribute significantly to safeguarding customer information. We catalogued these use cases and showed through both scholarly evidence and case examples that AI can detect anomalies and attacks that elude traditional tools, thereby preventing many data breaches or limiting their impact. We also confronted the challenges of adopting AI in this context, from technical issues like adversarial machine learning and data scarcity to operational and ethical concerns such as model explainability and regulatory compliance. A clear message is that while AI is a powerful ally, it is not a silver bullet – organizations must implement it within a robust governance and risk management framework. Strategies to mitigate these challenges (e.g., robust model training, human-in-the-loop oversight, privacy-preserving data handling, and alignment with compliance requirements) were discussed and should be part of any fintech’s AI deployment plan. Early adopters in the financial sector have demonstrated that these hurdles can be overcome through cross-disciplinary collaboration and iterative improvement . Importantly, we highlighted that regulatory bodies are acknowledging AI’s potential and are likely to require higher standards of care that AI can help meet, such as faster breach reporting and stronger authentication measures. Our experimental results provided strong quantitative backing to the argument that AI can markedly improve cybersecurity outcomes. The deep learning models we applied achieved detection rates above 98-99% on complex attack data, far outperforming legacy detection approaches. In practical terms, this means an AI-enhanced security operations center would catch almost every attempted intrusion or malicious activity, drastically reducing the window of opportunity for attackers to compromise consumer data. We showed how a Transformer-based detection system, for World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2819 instance, could have prevented or minimized several infamous breaches had it been in place, by recognizing the malicious patterns in real-time. Furthermore, we demonstrated that these models operate with high precision, ensuring that security teams can act on their alerts with confidence and without wasteful distraction from false alarms. The outcome is a virtuous cycle: more effective detection and response leads to fewer breached records and less frequent incidents, which in turn bolsters customer trust and eases regulatory pressure (since compliance metrics like breach frequency and response time improve). One of the most compelling findings is the dramatic reduction in breach detection and response times that AI enables. Our study, in line with industry reports, suggests that fintech firms using AI-driven security can identify breaches on the order of minutes or hours rather than days or months. This is transformative – the difference between a minor incident and a massive data leak often comes down to how quickly the attack is noticed and stopped. For consumers, this could mean that even if their data is targeted, the attack might be stopped before any significant data exfiltration occurs, or that they are notified almost immediately to take protective actions (like changing passwords) rather than finding out long after the damage is done. We conclude that integrating AI into fintech cybersecurity is not just beneficial but increasingly essential. As cyber threats continue to evolve in sophistication – with attackers themselves possibly using AI – traditional defenses will likely prove inadequate. AI provides the adaptability and learning ability needed to keep pace with dynamic threats . Fintech companies that leverage AI will be better positioned to protect their customers’ assets and information, comply with data protection regulations, and maintain a strong reputation for security. Conversely, those that do not adopt these advancements risk being outmaneuvered by attackers and losing consumer confidence. Our recommendations for fintech organizations and stakeholders are as follows: • Adopt AI-Driven Security Monitoring: Implement machine learning models (like those in this study) in critical security monitoring points – network traffic analysis, application log analysis, and transaction monitoring. Start with high-impact use cases such as intrusion detection and fraud detection, where mature solutions and models exist. Leverage open datasets and findings from academic research (such as the CIC-IDS datasets or published model architectures) as baselines to accelerate development. • Invest in Data and Training: Ensure collection of quality security data and continuously label and feed incidents back into the AI training process. Consider joining industry data-sharing initiatives (e.g., FS-ISAC) or using community breach databases (VCDB) to enrich training data while respecting privacy. Utilize techniques like federated learning if direct data sharing is not possible, so that models benefit from a wider range of threat examples. • Strengthen Governance around AI: Develop clear policies for AI usage in security. Define when AI alerts trigger automated actions versus human review. Incorporate AI into the incident response plan (e.g., “if AI flags critical severity, isolate the affected host immediately”). Establish oversight committees that include compliance and ethics roles to periodically review the AI system’s decisions for fairness and accuracy. Maintain documentation and audit trails for AI model updates and rationale for decisions to satisfy regulators and internal audit. • Address Explainability and Trust: Deploy tools to make AI decisions interpretable to analysts – for example, if a Transformer flags a transaction as fraud, provide the analyst with a feature importance or anomaly explanation (such as “deviates from user’s normal location and amount”). Training the security team to understand these models (perhaps with simplified mental models or visual aids) will increase trust and effective use. In parallel, educate executives and boards about the value and limitations of AI in mitigating cyber risk, so they allocate appropriate support and resources. • Regularly Evaluate and Update Models: Cyber threats evolve, and so must the AI models. Establish a cadence (e.g., quarterly) to retrain models with the latest data and threat intelligence. Use red teams to simulate novel attack techniques against the AI to identify blind spots (adversarial testing). Monitor model performance metrics over time in production – if false positives creep up or detection rates slip, investigate and retrain. By treating the AI models as living components of the infrastructure, fintechs can ensure they remain effective long-term. In terms of future work, as researchers we see the need for further studies on deploying these systems in real, live environments and measuring outcomes in situ (e.g., reduction in actual breach rates across companies using AI vs those not using it). Additionally, exploring more explainable and hybrid AI approaches (combining machine learning with rule-based logic) could yield systems that are both accurate and easier to regulate. Finally, as the arms race between attackers and defenders continues, research into adversarial robust AI for security will be crucial – ensuring that the next generation of attacks, possibly AI-assisted, can still be thwarted by resilient models. World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2820 In conclusion, the integration of artificial intelligence into fintech cybersecurity emerges from this study as a highly effective strategy for enhancing the protection of consumer data. With AI’s ability to detect threats swiftly and accurately, fintech companies can significantly reduce the likelihood and impact of data breaches and fraud. This not only protects consumers from financial loss and privacy violations but also strengthens the overall integrity and stability of the fintech ecosystem. As fintech services become ever more central to daily life, employing advanced AI-driven defenses will be integral to maintaining customer trust and securing the financial future. The path forward calls for embracing these technologies responsibly, informed by both technical evidence and governance considerations – a challenge that the fintech sector is well-poised to meet, as evidenced by the promising results and trends detailed in this work. Compliance with ethical standards Disclosure of conflict of interest No conflict of interest to be disclosed. References [1] Achuthan K, Ramanathan S, Srinivas S, Raman R. Advancing cybersecurity and privacy with artificial intelligence: current trends and future research directions. Front Big Data. 2024;7:Article 1497535. [2] Javaheri D, Fahmideh M, Chizari H, Lalbakhsh P, Hur J. Cybersecurity threats in FinTech: A systematic review. Expert Syst Appl. 2024; (In press). arXiv:2312.01752. [3] Oluokun A, Ige AB, Ameyaw MN. Building cyber resilience in fintech through AI and GRC integration: An exploratory study. GSC Adv Res Rev. 2024;20(1):228-237. [4] Dash N, Chakravarty S, Rath AK, Giri NC, AboRas KM, Gowtham N. An optimized LSTM-based deep learning model for anomaly network intrusion detection. Sci Rep. 2025;15(1):1554. [5] Ataa MS, Sanad EE, El-Khoribi RA. Intrusion detection in software-defined networks using deep learning approaches. Sci Rep. 2024;14(1):29159. [6] UpGuard. 10 Biggest Data Breaches in Finance. UpGuard Cyber Risk Blog. Jan 2025. [7] Secureframe (Emily Bonnie). 110+ of the Latest Data Breach Statistics [Updated 2025]. Secureframe Blog. Jan 2025. [8] Identity Theft Resource Center. 2023 Annual Data Breach Report. ITRC; Jan 2024. [9] Kroll. Data Breach Outlook: Finance Surpasses Healthcare as Most Breached Industry in 2023. Kroll Insights. Feb 2024. [10] SecurityWeek (Ionut Arghire). Finastra Starts Notifying People Impacted by Recent Data Breach. SecurityWeek News. Feb 18, 2025. [11] Equifax Data Breach Report. (Analysis of the 2017 Equifax Breach) – UpGuard Cyber Risk. 2019. [12] First American Financial Corp Breach Analysis. (UpGuard) 2019. [13] Treasury Department. Managing AI-Specific Cybersecurity Risks in the Financial Services Sector. US Dept. of Treasury Report. Oct 2023. [14] Sarker IH. AI in Cybersecurity: Identifying Evolving Threats with Adaptive Learning. J Inf Secur. 2023;14(2):115130. [15] Wewege L, Lee A, Thomsett C. AI in Finance: Applications in Fintech Security. Fintech Protect. 2020;5(3):33-41. [16] Kothandaraman B, Prasad A, Sivasankar E. AI-Driven SIEM for Financial Institutions. Proc. IEEE Int Conf CyberSec. 2023:112-119. [17] Arjunan R. Detecting Phishing Websites Using NLP and Deep Learning. IEEE Access. 2024; 12:10045-10056. [18] Despotović Z, Parmaković A, Miljković Z. Consequences of Data Breaches in Finance: Regulatory and Reputational Impact. J Financ Crime. 2023;30(1):128-142. World Journal of Advanced Research and Reviews, 2025, 26(01), 2802-2821 2821 [19] Vinuesa R, et al. The Role of Artificial Intelligence in Achieving the Sustainable Development Goals. Nat Commun. 2020; 11:233. [20] Apruzzese G, et al. Deep Learning for Anomaly Detection in Cybersecurity. IEEE Trans Neural Netw Learn Syst. 2022; (Early Access).