The healthcare industry benefits from electronic health records, telemedicine, telemonitoring, and IoMT, but these innovations have made it easier for attackers to infiltrate healthcare systems
[1]
. The ransomware attacks on Change Healthcare in 2024 and Universal Health Services in 2020 have shown the destructive potential of cyberattacks, which led to patient data loss, service downtime, and financial loss
[2]
[3]
. Healthcare providers are required to follow the HIPAA compliance requirements for data privacy and security, but they have been increasingly exposed to financial risks through fines, reputational loss, and even loss of patient care quality when their systems get compromised
[4]
.
Healthcare organizations invest heavily in cybersecurity solutions, but human-related vulnerabilities have emerged as the primary drivers of security breaches. The breach log review revealed that phishing attacks, together with credential theft, insider threats, and even security culture incidents, were the most reported type of incidents
[5]
. The academic frameworks CARE and C4 are human-centered frameworks for employee security, but they both lack the adaptive AI-driven learning capabilities to fight against dynamic cyber threats
[6]
[7]
. The cognitive gap in cybersecurity human vulnerabilities exists as these models are compliance-focused and awareness-driven, and they lack AI-driven adaptive capabilities. The current cyber defense technology provides effective cybersecurity, but human vulnerabilities still present exploitable threats which require anticipatory AI-driven human-centered solutions. A study published in Sensor’s journal showed how AI has been used to protect IoMT infrastructure through anomaly detection and prediction models for the prevention of healthcare data breaches
[8]
. The healthcare industry has been developing new applications of federated learning and blockchain technology for privacy-preserving decentralized security approaches in networked healthcare systems
[9]
. The recent advancements in adaptive artificial intelligence support the development of A4 AI-Guard as a human-centered security model. The solution to this challenge requires A4 AI-Guard as a new AI-driven threat defense system, which combines artificial intelligence with human-centered security approaches. The system A4 AI-Guard operates as a unified defense system, which integrates all four components instead of using the separate methods of predictive analytics, real-time threat detection, adaptive training, and cultural reinforcement. The system is built on the foundation of four adaptive pillars which form its infrastructure.
Main Objectives:
2. Literature Review
Healthcare cybersecurity requires a multidisciplinary approach, which brings together technological innovation with behavioral science. The current security frameworks CARE (Cognitive Awareness and Response Environment) and C4 (Cognitive Cybersecurity Framework) are primarily focused on the implementation of technical security controls and compliance requirements. The existing models have shown limited adaptability in the face of evolving human vulnerabilities
[6]
[7]
. Healthcare organizations are at major risk, as human elements, phishing, social engineering, and insider threats remain their weakest links
[5]
[12]
.
2.1. Cognitive/Behavioral Aspects
Studies in traditional cognitive psychology have shown that employee decision-making is impaired by overconfidence biases and confirmation biases, which result in security issues by facilitating cyber threats
[6]
[12]
. Healthcare organizations make incorrect risk judgments due to overconfidence, while their confirmation biases make them ignore security warnings that challenge their preexisting beliefs. The recognition of these aspects is essential for building cybersecurity models which guard against human vulnerabilities in stressful healthcare situations.
2.2. AI in Cybersecurity
The machine learning field has achieved impressive advances in computational algorithms by means of cognitively mutated algorithms, and this makes this work essential in the fight against both technical, human, and other vulnerabilities. Supervised learning algorithms, which consist of decision trees, support vector machines, and random forest algorithms, utilize labeled processed data to classify threats
[1]
. The unsupervised learning algorithms k-means clustering and autoencoders identify new attacks by isolating patterns that are dissimilar from normal user behavior
[11]
. The real-time optimization of the defensive moves by A4 AI-Guard is achieved through reinforcement learning (RL), which allows adaptivity for the protection against emerging threats and targeted attacks
[4]
[15]
.
The integration of neural networks with cross-validation and parameter tuning techniques improves system performance by minimizing false positives and at the same time enables transferability to different healthcare settings
[2]
[7]
. The predictive model based on historical attacks combines with the adaptive models based on streaming data to form a hybrid model that defends against both known and unknown threats
[1]
[5]
. Studies show that AI systems have been used to protect IoMT infrastructure through behavioral analysis, which identifies security threats before they mature into full-blown attacks. To provide context for these AI-driven tools, major industry evaluations have benchmarked the cybersecurity performance of healthcare organizations, identifying key areas of risk that require advanced solutions
[16]
. The proposed deep learning-based anomaly detection frameworks for IoMT healthcare systems have shown how predictive AI has enhanced active cyber defense capabilities
[17]
[18]
.
Research into federated learning and blockchain technology has continued to attract attention as these systems have shown promise for the protection of healthcare data privacy as well as enabling distributed network security. The study done by Abbas et al.
[9]
showed from their literature review that federated learning has the potential to protect patient data and at the same time share AI knowledge between healthcare facilities, and that blockchain technology can be used for secure data exchange in decentralized IoMT networks. The knowledge base reveals the need to construct A4 systems which support both adaptive and privacy-preserving operations.
2.3. Comparative Considerations
The majority of these efforts have been mostly technology-driven or compliance-oriented, and at the same time lacking human-centered defense features with AI-driven adaptivity
[13]
[19]
. As shown in
Table 1
, A4 AI-Guard provides stronger real-time threat detection compared to the other frameworks CARE and C4.
To address these limitations, A4 AI-Guard is proposed as a flexible, AI-enhanced defense system designed specifically for healthcare. It integrates machine learning, reinforcement learning, and cognitive insights to bridge the gap between technology and human vulnerabilities. As we describe in
Table 2
, the framework is structured around four adaptive pillars
[10]
:
Continuous AI monitoring of user behavior to detect anomalies and insider threats.
Active Threat Detection and Response
Real-time, AI-driven identification and containment of threats, leveraging adaptive learning to prevent the execution of malicious code.
Augmented Role-Based Training
Personalized training tailored to staff roles ensures situational preparedness against relevant risks.
Adaptive Cultural Reinforcement
AI analysis of compliance, incident reports, and feedback fosters a security-first organizational culture.
The four pillars function as an interdependent ecosystem. For instance, an anomaly detected by Adaptive Behavioral Risk Profiling can generate an automated alert to Active Threat Detection and Response, which in turn invokes Augmented Role-Based Training by issuing just-in-time learning prompts to relevant staff. Simultaneously, incident outcomes feed back to Adaptive Cultural Reinforcement to update organizational risk perception and resilience metrics. This closed feedback loop ensures that learning from one pillar continuously strengthens the others
[10]
.
As illustrated in
Figure 1
, the A4 AI-Guard framework operates as a closed-loop ecosystem in which each adaptive pillar reinforces the others to enhance healthcare cybersecurity.
<xref ref-type="bibr" rid="scirp.146620-"></xref>Figure 1. A4 AI Guard framework architecture.
The real-time interplay ensures that each pillar not only performs its specific role but also reinforces the others. For example, alerts from Adaptive Behavioral Risk Profiling can automatically trigger Active Threat Detection rules, prompt just-in-time learning through Augmented Role-Based Training, and feed cultural metrics back into Adaptive Cultural Reinforcement for continuous improvement.
2.5. Summary of Gaps
Upon completion of the literature search, three gaps were made evident based on the literature:
To address these gaps, this paper presents A4 AI-Guard as a scalable, proactive, and human-aware defense ecosystem for healthcare cyber.
3. The Method and Strategy
This study is best described as a conceptual-framework design validated through qualitative case-study analysis. Although we draw on a broad evidence base, the core contribution is the development of the A4 AI-Guard conceptual framework and its retrospective application to two major ransomware incidents
[14]
. This qualitative, case-anchored validation provides methodological rigor without implying quantitative mixed-method data collection.
The research method applied to the study has been rigorous, and the A4 AI-Guard model has a well-grounded theoretical and practical basis.
The initial phase of the process involved an extensive review of the available literature on Artificial Intelligence (AI) and its use in healthcare cybersecurity. The review also recorded an increasing call to make human-centric paradigms integrated, since technology-based solutions could also contribute to human errors, such as responses to phishing or social engineering
[12]
. The paper, therefore, accepted this insight and proposed A4 AI-Guard, a cybersecurity framework that bridges both the predictive and adaptive capabilities of AI with behavioral analytics to help provide a more comprehensive defense mechanism for health organizations.
The framework was then based on four major pillars that are adaptive in nature. They are: AI-backed behavioral risk profiling, Active Threat Detection and Response, Augmented Role-Based Training, and Adaptive Cultural Reinforcement, all backed by AI insights. The approach I have proposed through the A4 AI-Guard framework is designed to address all of these pillars. This, in turn, is a direct response to the gaps that have been evidenced in the current models in place, as it is easy to see that a tool like behavioral risk profiling is an alarm system that continuously observes employee activity and detects behavioral changes, such as abnormal logging in, accessing atypical data, etc., for early warning signs of a potential breach
[12]
. Also, using AI in a real-time threat detection framework can minimize incident response time by automating the process and allowing for early identification and even containment of attacks.
With augmented role-based adaptive training, relevant staff and employees can receive contextual education materials based on their roles and level of exposure in the organization, to help strengthen resilience at the organizational level. A more adaptive and dynamic model is preferred to traditional security awareness training, as the model can learn continuously based on feedback loops and reinforce behavior based on repeated anomalies. Culture reinforcement with AI insights and training, on the other hand, embeds a security cohesive culture that has an element of compliance as an actor’s everyday decision-making process. This approach to culture improvement has been further articulated regarding how healthcare organizations can intrinsically make their overall cybersecurity stance more robust for the long run, through continuing feedback loops and adaptation.
This human factor has been ensured to be expressed as the largest vulnerability factor and is still seen in healthcare cybersecurity environments. This postulation has been supported by the many case studies included in this paper, which show that human errors are still a large contributor to the overall success of the major ransomware attacks. As such, the A4 AI Guard system has a much better probability of predicting, preventing, and mitigating the occurrence of these types of risk, more comprehensively than a technology-only solution
[12]
.
In conclusion, improving the resilience factor within healthcare cybersecurity is dependent on the need to have adaptive, role-based training with cultural feedback, and real-time continuous monitoring systems to help meet the organizations’ objectives. The three pillars have all been previously discussed to make A4 AI Guard a living and adaptive framework, not only in terms of being able to respond to present-day attack threats, but also being more capable of defending against the next new class of threats as they arise. This, therefore, means that organizations can go to sleep every night knowing that they are hardening their defenses against all ongoing attacks, but also doing so for the long term by creating and embedding an element of learning and culture change.
4. Real-World Use Cases4.1. Case Study 1: Change Healthcare Ransomware Attack
In February 2024, Change Healthcare, one of the leading healthcare technology companies operating in the United States, experienced a large-scale ransomware attack
[22]
. Change Healthcare, whose operations for hospital pharmacies, help with obtaining medications, completing prescriptions, and supporting insurance processes was all in scramble mode when the company’s systems were impacted by the ransomware attack. The culprits for the ransomware were known to be AL-PHV/BlackCat, a notorious ransomware group that had stolen user credentials and gained access to their system
[2]
.
With their backdoor, the bad actors were able to successfully move laterally within the network and began to steal large amounts of sensitive patient and financial data and also began propagating ransomware throughout the system for the entire country. As a result, more than 190 million people were said to be affected by the attack. The attack had led to patients potentially being unable to receive important medications and/or book appointments, and the hospitals experiencing significant financial disruption that came from loss of reimbursements
[2]
. The total losses as a result of the incident were said to be hundreds of millions of dollars. This is if one were to take into account reputation loss and more when accounting for legal implications.
In such a case, if A4 AI-Guard had been in place, the outcomes would have been significantly reduced. The Adaptive Behavioral Risk Profiling module would have flagged behavioral anomalies, like strange log-ins and access to suspicious data, much sooner than what happened. Active Threat Detection and Response would have also responded much more quickly to lateral movement in the network and would most likely have been contained in a matter of hours rather than days or weeks
[3]
. Role-based Adaptive Training would have also ensured that the IT admins and even general IT support staff would have been able not just to detect but also to report and respond more quickly to prevent the escalation of the incident as a result of credential taking
[3]
. The last pillar, that of Adaptive Cultural Reinforcement, would have also ensured that there was a much more conscious effort made towards being more security cognizant and less of a response to critical data being leaked, as well as cultural conditions in which being less disruptive in day-to-day work might have contributed to the overall increase in the attack surface of the company.
4.2. Case Study 2: Universal Health Services (UHS) Ransomware Attack
In September 2020, Universal Health Services (UHS), the largest healthcare provider in the United States with over 400 hospitals and practices for mental and behavioral health, was the subject of a large ransomware attack
[23]
. The attack came in the form of a successful phishing email in which one of the employees clicked a nefarious link, dropped the malware, and then it spread across their internal systems
[2]
.
Within a few hours, all critical systems had been locked down, and the hospitals were made to operate as they had done many decades before—manually. Electronic Health Records (EHRs), emergency units, ambulatory service areas, and even surgical scheduling were all disabled, and even patient processing was difficult to achieve for a few days to weeks after the event. The disruptions were known to be so wide-reaching that this event has led to a recovery cost of over $67 million
[2]
.
In such a situation as this, had A4 AI-Guard been a policy or regulation in place, at least three distinct pillars of defensive countermeasures that we have discussed in this paper would have at least lessened the impact of the attack. Adaptive Behavioral Risk Profiling would have flagged abnormal email activities, which would then have been able to trigger a set of additional response measures, and also account for anomaly detection in the systems that were originally connected to the phishing effort. Active Threat Detection and Response would then have automatically kicked in to quarantine any affected endpoints, which would have saved the organization from having the entire infection in the first place.
[3]
. Augmented Role-Based Training would also have made certain all frontline staff would have been trained to a degree that would have included administrative, technical, nursing, and many other staff. Such training would have equipped all this frontline staff with skills to be able to easily identify phishing indicators or, better still, report the case before significant time had even elapsed and infection had spread.
[3]
. The last pillar of Adaptive Cultural Reinforcement is one that would also have ensured that there was a more substantive sense of security awareness in the organization as a whole, which would have made it less likely an employee would not have been able to recognize a ransomware phishing email and cause an organization-wide system compromise.
4.3. Summary
In conclusion, both cases help to show the vulnerability of healthcare organizations that are often open to compromise from a human component as well as cultural incompetency and more, with already-volatile and ever-changing cyber threats. In both cases, it is clear that gaps in human behaviors, system monitoring, and organizational readiness in these healthcare corporations would not have been compromised or leveraged if the adaptive AI-enabled A4 AI GUARD defenses had been deployed. A4 AI-Guard is, therefore, a reinforcement framework that addresses the gap by offering a more comprehensive and, of course, value-added model of enterprise defensible assurance approach that is clearly aligned to the four pillars that have been discussed earlier in the paper. It is also clear that the adoption of this new way of cybersecurity analysis, via its predictive analytics and real-time detection, tailored training, and adaptive cultural reinforcement, has the capacity to help organizations reduce, at the very least, the impact of such ransomware or data-theft attacks or, at best, ensure their systems are not compromised, and everything from capacity, delivery, financial, and reputational damage has been considered or, in this case, already realized.
Table 3
provides a side-by-side comparison of the two ransomware incidents, summarizing the exploited vulnerabilities, resulting impacts, and how A4 AI-Guard’s four adaptive pillars could have mitigated each attack.
<xref ref-type="bibr" rid="scirp.146620-"></xref>Table 3. Case study comparison.
No
Incident
Vulnerability Exploited
Impact
How A4 AI-Guard Would Mitigate
1
Change Healthcare (2024)
Stolen credentials → Undetected lateral movement
190M patients affected, nationwide service disruption, $100M+ in costs
Behavioral Profiling flags anomalies; Real-time Detection stops the spread; Role-based training prevents credential theft
2
UHS (2020)
Phishing email → Malware propagation
400 hospitals disrupted, EHR down, $67M+ recovery costs
Phishing detection; auto-quarantine infected endpoints; role-based training for staff; cultural reinforcement reduces risk
5. Cost-Benefit Analysis
The estimated costs of A4 AI-Guard implementation are based on industry benchmarks and real-world cases of large-scale enterprise cybersecurity ecosystem rollouts incorporating advanced AI-based technologies, primarily within the healthcare sector. Because no practical implementation of A4 AI-Guard has been undertaken as of yet, these estimates are based on the best-practice models and related technologies of extant AI-based cybersecurity technologies. System Integration Costs:
Deploying AI-driven cybersecurity applications typically represents a considerable investment in the hardware, software, and professional services required to customize them for the organization’s needs and the AI-driven incident tracking workflows. In total, industry reports and case studies point to the costs of integration being within a range of $200,000 to $500,000. This includes the associated costs of deploying machine-learning infrastructure and data pipelines and, in the case of cloud-specific solutions, the costs of the cloud services. The costs will, of course, vary based on the size of the healthcare facility, the state of its existing IT systems, and the amount and complexity of AI Model(s)
[24]
[25]
.
Employee training is a very important component to ensuring that your AI-driven cybersecurity framework is suitably integrated into your organizational processes and strategic direction. Notably, the development of specified training programs and attendance at training sessions by employees depends on the maturity of the overall project. The overall costs for overall role-based training programs are estimated at a range of $50,000 to $100,000 based on prevailing industry practice for large healthcare organizations. This includes the necessary costs to develop the training modules, expenses for hiring subject-matter experts/consultants, as well as the payer’s costs for the employee attending training sessions where they will be paid during the training time
[26]
. Spending on Augmented Role Based Training, as an example, will reflect direct staff training, which reflects a clear, concrete, and measurable return on investment as staff develop knowledge unique to their organizational role.
AI systems require ongoing updates, patching, and optimizations to remain effective in defending against the dynamic landscape of cyber threats. The estimated continuing cost of maintaining these systems is $5,000 a month for cloud infrastructure maintenance, occasional upgrades, and system optimization
[21]
. The frequency of changes in the healthcare threat landscape necessitates that these ongoing costs are incurred to sustain the adaptability and resilience of A4 AI-Guard.
To highlight the economic feasibility of the proposed framework,
Figure 2
presents the cost-benefit comparison for A4 AI-Guard implementation, demonstrating how the anticipated long-term savings and risk-reduction benefits outweigh the initial deployment and maintenance costs.
<xref ref-type="bibr" rid="scirp.146620-"></xref>Figure 2. Presents the cost-benefit comparison for A4 AI-Guard implementation.
6. Discussion and Implications
A4 AI-Guard is a substantial step forward in healthcare cybersecurity. Action4 is a frame of reference that legitimizes advancing beyond the compliance-based approaches to a (preemptive) preparedness model for cyber-attacks utilizing AI, and as such, can have benefits in enhancing the overall system’s defensive infrastructure. The A4 AI-Guard framework intentionally recognizes that while technology is important, cybercriminals target human behaviour and organizational culture most often
[13]
.
Collectively, these four adaptive pillars act as a single, interdependent ecosystem, where detection events can automatically drive real-time responses, trigger targeted staff training, and feed cultural metrics back into the system for continuous improvement.
6.1. Benefits
The framework offers healthcare organizations a wide array of pragmatic benefits, including:
6.2. Implications
The potential implications of A4 AI-Guard are significant and go beyond facilitating its implementation in healthcare.
7. Limitations and Future Work
Despite the above advantages, there are several limitations of A4 AI-Guard to acknowledge. The framework is heavily dependent on a behavioral and system data set that is of sufficiently high quality, which means that bias, incompleteness, or noise in the inputs can compromise detection. Additionally, the ongoing evolution of healthcare cyber threats presents a continual need for updates, retraining AI models, and optimizing systems. AI transparency and explainability have yet to be resolved—the less clinicians and IT trust any sort of black box prediction, the less they will rely on its recommendations. The cost of integration and organizational cultural adoption may also become a hindrance for smaller healthcare providers with inherently limited funding and resources
[29]
.
Public safety, privacy, and data security are also paramount. Because healthcare data is especially sensitive, there is a significant need for compliance with HIPAA, GDPR, and related regulations. Advanced methods, including federated learning, offer collaborative training between organizations and institutions without having to centralize data, and blockchain offers a secure and auditable way to share medical records. While a recent review done by Abbas et al.
[9]
highlights the potential for combining these approaches to create privacy-preserving, decentralized security infrastructures.
Effective operation requires high-quality, multimodal data, such as EHR access logs, network flow records, endpoint telemetry, and user-activity metadata. While these inputs are vital for accurate detection and behavioral modeling, they raise significant privacy and regulatory concerns. Our design anticipates mitigations such as federated learning to keep raw patient data local, differential privacy in model training, and HIPAA/GDPR-compliant access controls to ensure lawful processing.
Future work should focus on the piloting of A4 AI-Guard in actual healthcare environments to assess its functionality in real time, and to benchmark against currently used cybersecurity frameworks. To do this effectively, XAI (explainable AI) methods will need to be incorporated to promote trust and buy-in from healthcare stakeholders. It will also be important to validate the scalability of A4 AI-Guard in different healthcare ecosystems—from small practices, to smaller networks, to larger national hospital networks—to assess its longer-term viability.
This study outlined a robust, flexible, human-centered cybersecurity solution for healthcare in the context of an evolving cyber threat environment, but there are factors that must be addressed to ensure the appropriate technology solution can be deployed that encompasses the emerging technology as discussed and designed in earlier sections, to promote visible adaptation and mitigation of AI, cybersecurity, and human behaviors on a human-centered adaptive model with significant reinforcement to improve healthcare cyber resilience.
8. Conclusions
This rise in cyber threats provides an opportunity to seek solutions that focus on both human vulnerabilities and technological vulnerabilities to offer real risk mitigation for the healthcare system as a whole. This paper proposes A4 AI-Guard, which enables a comprehensive understanding and perspective of cybersecurity with a human focus, using artificial intelligence and human-centered approaches to address human vulnerabilities. A4 AI-Guard is not an average approach to cybersecurity; instead of focusing solely on AI (artificial intelligence) perspectives, A4 AI-Guard incorporates four adaptive pillars to deliver a holistic and evolving open approach to targeting ransomware, phishing, insider threats, and other methods aimed at healthcare or in the healthcare industry through adaptive behavioral risk profiling, active threat detection and incident response, augmented role-based development, and adaptive cultural reinforcement.
Using the Change Healthcare
[22]
and Universal Health Services
[23]
ransomware situations, this paper also highlighted what A4 AI-Guard could have done to limit the scope and impact of the incidents brought to the owner’s attention by providing reduced detection latency, limiting lateral movement, and developing ready, willing, and able staff. A4 AI-Guard’s reliance on systemic and time-driven continuous AI adaptability for recognition/training and its capacity to right-size the cultural reinforcement to be feasible and memorable in a dynamic, trivially impacting, and complex healthcare environment lends itself as a viable method to drive systemic adaptability.
Future validation of A4 AI-Guard through real implementation is a next step, and further empirical testing should be carried out through piloting the A4 AI-Guard in varying healthcare environments, benchmarking with existing cybersecurity frameworks, and further exploring the addition of blockchain and federated learning, data sharing, and further privacy of underlying AI models. As well, addressing the challenges to mitigate AI concerns (bias, etc.) to ensure privacy in workloads and ensuring security possesses scalability will be a limitation to be aware of when looking to validate the practical uptake when looking to deploy the A4 AI-Guard as a solution.
In conclusion, A4 AI-Guard is an adaptable, scalable, hierarchical, proactive, and human-centric defense ecosystem. Through the merged approaches of AI and further humanization of behavioral observation and cultural transformational changes, A4 AI-Guard seeks to enable healthcare organizations to develop the resilience to survive the evolving nature of cyberthreats that are becoming increasingly more complex.
References
He, Y., Aliyu, A., Evans, M. and Luo, C. (2021) Health Care Cybersecurity Challenges and Solutions under the Climate of COVID-19: Scoping Review. Journal of Medical Internet Research, 23, e21747.
>https://doi.org/10.2196/21747
Shryock, T. (2024) Ransomware Surge Highlights Critical Cybersecurity Gaps in Health Care. >https://www.medicaleconomics.com/view/ransomware-surge-highlights-critical-cybersecurity-gaps-in-health-care
Alder, S. (2025) Change Healthcare Increases Ransomware Victim Count to 192.7 Million Individuals.
>https://www.hipaajournal.com/change-healthcare-responding-to-cyberattack/
Zhan, Y., Ahmad, S.F., Irshad, M., Al-Razgan, M., Awwad, E.M., Ali, Y.A., et al. (2024) Investigating the Role of Cybersecurity’s Perceived Threats in the Adoption of Health Information Systems. Heliyon, 10, e22947.
>https://doi.org/10.1016/j.heliyon.2023.e22947
Burke, W., Stranieri, A., Oseni, T. and Gondal, I. (2024) The Need for Cybersecurity Self-Evaluation in Healthcare. BMC Medical Informatics and Decision Making, 24, Article No. 133.
>https://doi.org/10.1186/s12911-024-02551-x
Rahmany, M. and Selvi, A. (2025) C4 Framework for Healthcare Cybersecurity Defense: A Human-Centric, Socio-Technical Approach. E-Health Telecommunication Systems and Networks, 14, 31-38.
>https://doi.org/10.4236/etsn.2025.143004
Akhtar, Z.B. and Tajbiul Rawol, A. (2024) Enhancing Cybersecurity through Ai-Powered Security Mechanisms. IT Journal Research and Development, 9, 50-67.
>https://doi.org/10.25299/itjrd.2024.16852
Yigit, Y., Ferrag, M.A., Ghanem, M.C., Sarker, I.H., Maglaras, L.A., Chrysoulas, C., et al. (2025) Generative AI and LLMs for Critical Infrastructure Protection: Evaluation Benchmarks, Agentic AI, Challenges, and Opportunities. Sensors, 25, Article 1666.
>https://doi.org/10.3390/s25061666
Abbas, S.R., Abbas, Z., Zahir, A. and Lee, S.W. (2024) Federated Learning in Smart Healthcare: A Comprehensive Review on Privacy, Security, and Predictive Analytics with IoT Integration. Healthcare, 12, Article 2587.
>https://doi.org/10.3390/healthcare12242587
Balla, E. (2025) Pillars of Intelligence: Key Fields at the Forefront of AI.
>https://aijourn.com/pillars-of-intelligence-key-fields-at-the-forefront-of-ai/
Gupta, S., Kapoor, M. and Debnath, S.K. (2025) AI-Based Cybersecurity Solutions for Healthcare. In: Artificial Intelligence-Enabled Security for Healthcare Systems, Springer, 65-86.
>https://doi.org/10.1007/978-3-031-82810-2_4
Sudheer, S. (2024) Ransomware Attacks and Their Evolving Strategies: A Systematic Review of Recent Incidents. Journal of Technology and Systems, 6, 32-59.
>https://doi.org/10.47941/jts.2399
Javaid, M., Haleem, A., Singh, R.P. and Suman, R. (2023) Towards Insighting Cybersecurity for Healthcare Domains: A Comprehensive Review of Recent Practices and Trends. Cyber Security and Applications, 1, Article 100016.
>https://doi.org/10.1016/j.csa.2023.100016
Universal Health Services Inc (2020) Form 8‑K: Information Technology Security Incident. U.S. Securities and Exchange Commission.
>https://www.sec.gov/Archives/edgar/data/352915/000156459020044863/uhs-8k_20200927.htm
Binhammad, M., Alqaydi, S., Othman, A. and Abuljadayel, L.H. (2024) The Role of AI in Cyber Security: Safeguarding Digital Identity. Journal of Information Security, 15, 245-278.
>https://doi.org/10.4236/jis.2024.152015
Censinet, K.L.A.S., American Hospital Association, Health‑ISAC and HPHSCC (2025) 2024 Healthcare Cybersecurity Benchmarking Study.
>https://health-isac.org/partnered-report-healthcare-cybersecurity-benchmarking-study-2024/
International Research Journal of Multidisciplinary Studies (2025) Deep Learning for Anomaly Detection in IoT Healthcare Systems.
>https://www.irjms.com/journal/deep-learning-for-anomaly-detection-in-iot-healthcare-systems
Alalwany, E., Alsharif, B., Alotaibi, Y., Alfahaid, A., Mahgoub, I. and Ilyas, M. (2025) Stacking Ensemble Deep Learning for Real-Time Intrusion Detection in IoMT Environments. Sensors, 25, Article 624.
>https://doi.org/10.3390/s25030624
Ewoh, P. and Vartiainen, T. (2024) Vulnerability to Cyberattacks and Sociotechnical Solutions for Health Care Systems: Systematic Review. Journal of Medical Internet Research, 26, e46904.
>https://doi.org/10.2196/46904
Neprash, H.T., McGlave, C.C., Cross, D.A., Virnig, B.A., Puskarich, M.A., Huling, J.D., et al. (2022) Trends in Ransomware Attacks on US Hospitals, Clinics, and Other Health Care Delivery Organizations, 2016-2021. JAMA Health Forum, 3, e224873.
>https://doi.org/10.1001/jamahealthforum.2022.4873
Mishra, S. (2023) Exploring the Impact of AI-Based Cyber Security Financial Sector Management. Applied Sciences, 13, Article 5875.
>https://doi.org/10.3390/app13105875
Pults, K. (2025) The Change Healthcare Breach: What Changed, What Didn’t, and What Must.
>https://www.securitymagazine.com/articles/101394-the-change-healthcare-breach-what-changed-what-didnt-and-what-must
Newman, L. (2020) A Ransomware Attack Has Struck a Major US Hospital Chain.
>https://www.wired.com/story/universal-health-services-ransomware-attack
IBM (2025) Cost of a Data Breach: The Healthcare Industry.
>https://www.ibm.com/think/insights/cost-of-a-data-breach-healthcare-industry
National Cyber Security Centre (NCSC) (2025) Annual Review 2024.
>https://www.ncsc.gov.uk/files/NCSC_Annual_Review_2024.pdf
Ponemon Institute (2024) The 2024 Study on Cyber Insecurity in Healthcare: The Cost and Impact on Patient Safety and Care.
>https://ponemonsullivanreport.com/2024/10/the-2024-study-on-cyber-insecurity-in-healthcare-the-cost-and-impact-on-patient-safety-and-care/
Health‑ISAC (2024) Partnered Report: Healthcare Cybersecurity Benchmarking Study 2024.
>https://health-isac.org/partnered-report-healthcare-cybersecurity-benchmarking-study-2024/
Mennella, C., Maniscalco, U., De Pietro, G. and Esposito, M. (2024) Ethical and Regulatory Challenges of AI Technologies in Healthcare: A Narrative Review. Heliyon, 10, e26297.
>https://doi.org/10.1016/j.heliyon.2024.e26297
Asan, O., Bayrak, A.E. and Choudhury, A. (2020) Artificial Intelligence and Human Trust in Healthcare: Focus on Clinicians. Journal of Medical Internet Research, 22, e15154.
>https://doi.org/10.2196/15154