AI-augmented workflow resilience framework for cybersecurity risk mitigation in hospital AI systems

 Corresponding author: HARI SURESH BABU GUMMADI. Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution License 4.0. AI-augmented workflow resilience framework for cybersecurity risk mitigation in hospital AI systems HARI SURESH BABU GUMMADI * Jawaharlal Nehru Technological University, Hyderabad, India. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 Publication history: Received on 29 March 2025; revised on 06 May 2025; accepted on 09 May 2025 Article DOI: https://doi.org/10.30574/wjarr.2025.26.2.1754 Abstract The AI-Augmented Workflow Resilience Framework represents a transformative approach to cybersecurity in healthcare environments utilizing artificial intelligence systems. It examines how the integration of AI into hospital settings creates unique security vulnerabilities that traditional cybersecurity methods fail to adequately address. The proposed framework embeds security mechanisms directly into clinical and administrative workflows through five interconnected layers: Continuous Workflow Monitoring, AI-Specific Threat Detection, Healthcare Context Interpretation, Adaptive Response Orchestration, and Continuous Learning and Improvement. Implementation across diverse healthcare facilities—including community hospitals, regional medical centers, and academic medical centers— demonstrated the framework's effectiveness in enhancing security while preserving operational efficiency. Evaluation results reveal substantial improvements in threat detection capabilities, particularly for AI-specific vulnerabilities such as adversarial attacks and model manipulation. The context-aware approach significantly reduced false positives and workflow disruptions while maintaining essential clinical functions during security incidents. Technical performance analysis confirmed reasonable resource requirements with favorable scalability characteristics. It addresses a critical gap in healthcare cybersecurity by creating an integrated approach that protects advanced AI systems while supporting rather than impeding the delivery of patient care. Keywords: Healthcare Cybersecurity; Artificial Intelligence Security; Workflow Resilience; Context-Aware Security; Adversarial Attack Detection 1. Introduction The integration of artificial intelligence into healthcare environments presents both revolutionary opportunities and unprecedented security challenges. Modern healthcare facilities now deploy AI systems across a spectrum of functions, from administrative process automation to sophisticated clinical decision support tools. A comprehensive analysis of healthcare institutions revealed that a significant majority had implemented at least one AI system by 2023, with multiple distinct AI applications per facility. These implementations have demonstrated substantial benefits, including reduction in administrative processing times and improvement in diagnostic accuracy for specific conditions. However, this digital transformation has simultaneously created novel attack vectors that conventional security approaches are inadequately prepared to address. A systematic review of healthcare cybersecurity incidents between 2020-2023 documented numerous reported breaches affecting millions of patient records, with AI-related vulnerabilities implicated in a considerable percentage of cases—a figure that has increased at a substantial annual rate during this period. These statistics highlight the urgent need for specialized security frameworks tailored to the unique challenges of AI-enabled healthcare environments. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1176 1.1. Healthcare AI Security Landscape The security challenges in healthcare AI systems stem from their distinct operational context. Healthcare organizations face mounting pressure to simultaneously embrace transformative technologies while protecting highly sensitive patient data and maintaining uninterrupted service availability. The Healthcare Information Security Survey conducted across hundreds of healthcare institutions identified that a majority of facilities experienced at least one AI-related security incident in 2023, with significant remediation costs per breach and considerable mean downtime. The survey further revealed significant capability gaps, with only a small percentage of healthcare organizations reporting high confidence in their ability to detect AI-specific security threats such as adversarial attacks or data poisoning attempts. Traditional security tools detected only a fraction of AI-related security incidents during controlled penetration tests, demonstrating the inadequacy of conventional approaches for these emerging threat vectors. Among the most vulnerable AI applications were clinical decision support systems, automated documentation systems, and resource allocation algorithms, with consequences ranging from data exposure to clinical recommendation manipulation and operational disruption. These findings underscore the need for healthcare-specific security approaches that address the unique challenges introduced by AI integration. 2. The ai-augmented workflow resilience framework The proposed framework represents a paradigm shift from conventional cybersecurity approaches by embedding security mechanisms directly into clinical and administrative workflows rather than implementing them as isolated protective layers. This integration enables simultaneous optimization of both security posture and operational efficiency through five interconnected functional layers: 2.1. Continuous Workflow Monitoring Layer This foundational component establishes comprehensive behavioral baselines for both AI and human elements within healthcare workflows. Unlike conventional monitoring approaches that focus on network traffic or system logs, this layer implements continuous process mining that captured a high percentage of workflow variations across multiple clinical departments during validation studies. The monitoring system employs temporal pattern recognition that reduces false positive rates considerably compared to traditional anomaly detection methods by distinguishing between legitimate procedural variations and potential security threats. Implementation across academic medical centers demonstrated that this approach successfully identified most simulated attack scenarios while generating acceptable alert volumes that did not overwhelm security personnel. A key innovation is the integration of clinical context detection that automatically recognized legitimate emergency protocols that would have triggered false alarms in conventional security systems. Table 1 Framework Layers [2] Layer Primary Function Key Benefits Workflow Monitoring Establish baseline patterns Anomalous deviation detection, Reduced false positives AI-Specific Threat Detection Identify AI-targeted attacks Protection against adversarial inputs and model tampering Context Interpretation Provide clinical context Appropriate alert prioritization, Reduced disruptions Response Orchestration Coordinate security responses Maintained clinical functions during incidents Continuous Learning Improve over time Enhanced detection, Proactive protection 2.2. AI-Specific Threat Detection Layer Building upon the workflow monitoring foundation, this layer implements specialized mechanisms addressing vulnerabilities unique to AI systems. Validation testing against a standardized threat library demonstrated high detection rates for adversarial inputs, data poisoning attempts, and unauthorized model modifications. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1177 Research utilizing thousands of medical images with various adversarial perturbations showed that adversarial attacks successfully manipulated diagnostic AI to produce incorrect results in a majority of cases when using targeted perturbations below human perception thresholds. The implemented detection mechanisms reduced this successful manipulation rate significantly through multi-layered verification that combines statistical pattern analysis, model consistency validation, and automated secondary verification for high-risk decisions. A particularly effective component was the model integrity verification system that created cryptographic model signatures and continuously validated execution patterns, detecting nearly all attempted model tampering events during red team exercises while adding minimal latency to inference operations. This minimal performance impact proved crucial for maintaining clinical workflow efficiency. 2.3. Healthcare Context Interpretation Layer The effectiveness of security responses in healthcare environments depends critically on understanding the operational context in which potential threats emerge. This layer incorporates domain-specific knowledge to properly prioritize responses based on clinical impact. Evaluation across clinical departments demonstrated that the context-aware system correctly classified a high percentage of anomalous events according to their potential impact on patient care, compared to a much lower correct classification rate using generic security frameworks. The system employs a healthcare-specific knowledge base containing numerous distinct clinical workflows with validated procedural variations that enable it to distinguish between legitimate operational adaptations and security concerns. This context-awareness reduced unnecessary workflow disruptions significantly compared to traditional security approaches during operational evaluation. A notable innovation is the patient acuity-aware prioritization that dynamically adjusts security responses based on clinical urgency, implementing less disruptive measures in critical care environments while maintaining protection. This approach decreased care disruptions while maintaining most security effectiveness during simulated incidents. 2.4. Adaptive Response Orchestration Layer When threats are detected, this layer coordinates appropriate responses while preserving essential clinical functions. Unlike traditional security approaches that often implement binary allow/block decisions, the adaptive response system employs graduated interventions calibrated to both threat severity and clinical context. The orchestration layer implements various distinct response patterns ranging from enhanced monitoring to full component isolation, selecting optimal interventions based on threat characteristics and operational impact. Implementation across healthcare systems demonstrated that this approach maintained most critical functionality during active security incidents compared to considerably less functionality preservation with conventional security controls. A distinctive feature is the workflow reconfiguration capability that automatically established alternative processing paths for interrupted functions during containment procedures, ensuring continuity of essential services. The system's resource reallocation protocols successfully redirected clinical workflows around compromised components much faster than manual intervention procedures. 2.5. Continuous Learning and Improvement Layer The framework incorporates adaptive mechanisms that continuously enhance its effectiveness based on operational experience and emerging threats. A federated learning system aggregated anonymized security insights across multiple participating healthcare organizations, enabling identification of novel attack patterns substantially earlier than isolated monitoring systems. The near-miss analysis module identified many potential vulnerabilities that had not manifested as security incidents during an evaluation period, allowing preemptive remediation. Simulation-based training automatically generated numerous attack scenarios to improve detection algorithms, increasing identification rates for novel threats compared to static rule-based approaches. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1178 2.6. Implementation Outcomes A comprehensive evaluation across major hospital systems demonstrated significant security and operational improvements. The framework reduced successful adversarial attacks against clinical decision support systems substantially while decreasing the mean time to detect AI-specific security incidents from many hours to just a few. Importantly, critical services maintained near-total availability during active security incidents compared to a lower baseline with conventional security controls. The false positive rate for security alerts decreased significantly, reducing workflow disruptions. Staff satisfaction with security measures improved by a notable percentage based on standardized surveys, largely attributed to the reduction in unnecessary interventions that had previously impeded clinical work. The economic impact assessment indicated a substantial reduction in overall security incident costs, with the most significant savings in reduced downtime and decreased remediation expenses. The return on investment calculation yielded a positive financial outcome within months of implementation. The AI-Augmented Workflow Resilience Framework represents a significant advancement in healthcare cybersecurity by addressing the unique challenges posed by AI integration. By embedding security directly into clinical workflows rather than treating it as a separate function, healthcare organizations can simultaneously enhance their security posture while maintaining operational efficiency. The framework's context-aware approach ensures that security measures complement rather than impede the delivery of patient care, addressing a critical limitation of conventional cybersecurity approaches in healthcare environments. 3. Implementation and Evaluation of the AI-Augmented Workflow Resilience Framework 3.1. Implementation Approach The AI-Augmented Workflow Resilience Framework was deployed across three distinct healthcare environments: a community hospital, a regional medical center, and an academic medical center. This stratified implementation approach provided comprehensive insights into the framework's adaptability across diverse clinical settings and technological infrastructures. According to detailed implementation analysis, the complete integration process required several months on average, with larger facilities generally requiring longer deployment timeframes due to their more complex technological ecosystems [5]. The implementation followed a structured four-phase methodology that minimized disruption to critical healthcare operations while ensuring comprehensive security coverage. Table 2 Implementation Sites [5] Facility Type Size Implementation Duration Primary AI Applications Community Hospital 120 beds 3 months Clinical documentation, Medication management Regional Medical Center 450 beds 4.5 months Decision support, Resource allocation Academic Medical Center 800 beds 5 months Advanced diagnostics, Predictive analytics The initial baseline establishment phase lasted several weeks and focused on deploying monitoring components to map existing workflows and establish operational baselines. This phase revealed that a significant majority of healthcare IT systems required custom API development to enable proper security integration, highlighting the heterogeneous nature of healthcare technology environments [5]. Process mining techniques during this phase captured a substantial proportion of medical devices and systems within the monitoring scope, providing comprehensive visibility across the technology landscape [6]. Initial security monitoring during this baseline phase identified numerous potential security events across the monitored environments, establishing a preliminary threat profile for each institution that informed subsequent implementation phases. The integration phase extended over several weeks and connected security components with existing clinical workflows and systems. Implementation teams invested substantial person-hours per facility during this phase, with effort requirements scaling approximately linearly with institutional size [5]. Integration challenges were most pronounced with legacy systems, particularly medical devices with limited security functionality. Monitoring coverage assessments revealed successful integration with the vast majority of medical devices across implementation sites, with the remaining devices requiring compensating controls due to technical limitations [6]. Throughout this phase, existing World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1179 security systems operated in parallel to ensure continuous protection, creating a valuable comparison environment for subsequent effectiveness evaluation. The calibration phase lasted approximately a month and focused on tuning detection thresholds and response mechanisms. Initial monitoring revealed a concerning rate of false positives, which was subsequently reduced substantially through algorithmic optimization and contextual rule refinement [6]. This calibration phase proved particularly challenging for AI-specific threat detection, as these novel threat vectors lacked extensive historical data for pattern recognition. The implementation team documented that many healthcare institutions initially failed to meet all regulatory requirements for AI systems security, necessitating substantial configuration adjustments during this phase [7]. Detection tuning placed particular emphasis on identifying ransomware, data exfiltration attempts, and unauthorized access, reflecting the most common threat types observed during baseline monitoring. The operational phase established ongoing monitoring and continuous improvement mechanisms. Postimplementation analysis documented significant improvement in data processing efficiency and a notable reduction in IT operational costs over the months following full deployment [5]. These efficiency gains resulted primarily from reduced manual security monitoring requirements and decreased incident response burdens due to improved threat prevention capabilities. The framework's adaptive learning mechanisms demonstrated consistent monthly improvement in prediction accuracy through continuous operation, reflecting the systems' ability to refine detection algorithms based on operational experience [9]. 3.2. Evaluation Methodology The framework underwent rigorous evaluation across security effectiveness, operational impact, and technical performance dimensions. The evaluation methodology incorporated both quantitative metrics and qualitative assessments to provide comprehensive performance insights. Security effectiveness testing utilized controlled redteam exercises simulating various attack scenarios, while operational impact assessment combined workflow monitoring with user experience surveys. Technical performance evaluation focused on resource utilization, scalability, and integration requirements across the implementation sites. Security effectiveness evaluation employed controlled simulations of multiple attack vectors, with particular emphasis on AI-specific threats that traditional security measures often fail to detect. Threat simulation included adversarial manipulation attempts targeting clinical decision support systems, data poisoning attacks against learning algorithms, and unauthorized model manipulation efforts. Throughout the evaluation period, AI-related security incidents showed a concerning year-over-year increase across the healthcare sector, underscoring the importance of specialized protection mechanisms [7]. The red-team exercises were conducted at regular intervals over several months, with attack sophistication increasing progressively to simulate adaptive adversaries and test the framework's learning capabilities. Operational impact assessment utilized a mixed-methods approach combining quantitative workflow metrics with qualitative staff feedback. Continuous monitoring documented workflow interruption frequency, duration, and clinical impact before and after implementation. User experience surveys revealed that a majority of healthcare professionals initially expressed concerns about AI data handling practices, highlighting the importance of addressing both security and privacy considerations in the framework design [7]. Follow-up assessments after implementation documented that most clinicians reported workflow improvements with the context-aware security approach, reflecting the framework's success in balancing security requirements with operational efficiency [9]. Particular attention was paid to high-acuity clinical areas where workflow disruptions could have significant patient safety implications. Technical performance evaluation focused on computational overhead, scalability characteristics, and maintenance requirements. Detailed performance monitoring collected system resource utilization metrics throughout the evaluation period, providing insights into the framework's efficiency. Anomaly detection algorithms demonstrated high accuracy in identifying security threats, with false positive rates decreasing substantially through contextual refinement [8]. The framework's resource requirements showed linear scaling relative to institutional size and transaction volume, enabling accurate capacity planning for future implementations. Integration effort and ongoing maintenance requirements were meticulously documented, with maintenance needs decreasing over time as the system's learning mechanisms reduced manual tuning requirements. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1180 3.3. Security Effectiveness Results The framework demonstrated significant improvements in threat detection and response capabilities compared to traditional security approaches. Anomaly detection algorithms achieved high accuracy in identifying potential threats, substantially outperforming conventional signature-based approaches for similar threat profiles [8]. This improvement was particularly pronounced for AI-specific threats such as adversarial attacks and data poisoning attempts, which traditional security measures frequently failed to identify. The average detection time for critical threats was markedly faster compared to conventional security systems operating in parallel [6]. Table 3 Security Effectiveness [6] Threat Type Traditional Security Framework Performance Ransomware Moderate detection, Delayed response High detection, Rapid response Adversarial AI Manipulation Very low detection High detection, Effective prevention Data Poisoning Low detection High detection, Early intervention Unauthorized Access Moderate detection, High false positives High detection, Low false positives Model Tampering Very low detection High detection, Minimal latency impact Ransomware detection capabilities proved especially valuable, with the framework identifying most simulated ransomware attacks during early preparation stages before significant encryption could occur. Early detection enabled preemptive intervention that prevented operational impacts in most simulation scenarios. The context-aware detection algorithms proved particularly effective at identifying anomalous file access patterns indicative of ransomware activity while correctly distinguishing these from legitimate clinical access patterns. This contextual awareness reduced false positives considerably compared to traditional approaches that lacked healthcare-specific operational context [9]. The framework's specialized mechanisms for detecting adversarial manipulations of AI systems demonstrated substantial effectiveness, identifying a high percentage of subtle input manipulations targeting clinical decision support systems. These detection capabilities addressed a critical vulnerability in healthcare AI deployments, particularly for diagnostic systems where manipulation could lead to incorrect clinical recommendations. The implementation of AIspecific protections resulted in a substantial reduction in successful attacks against AI components compared to baseline security configurations [7]. Detection effectiveness varied by attack sophistication, with higher success rates against basic attack methodologies compared to advanced techniques employing generative AI for attack vector development. The system's ability to identify model drift and gradual degradation of AI performance proved particularly valuable for maintaining clinical safety. The framework successfully detected subtle model performance changes well before they became clinically significant, enabling proactive model retraining or verification. This early detection leveraged statistical pattern analysis that identified shifts in model output distributions that preceded observable accuracy declines. The context-aware monitoring achieved high accuracy in recognizing clinically significant drift patterns while correctly dismissing normal operational variations [9]. 3.4. Operational Impact Results Implementation of the framework demonstrated positive effects on workflow efficiency and operational continuity across all evaluation sites. The integration of security mechanisms directly into clinical and administrative workflows enabled simultaneous optimization of security and operational efficiency, addressing a critical limitation of traditional security approaches that often impose significant workflow burdens. The context-aware security approach reduced false positives substantially compared to conventional methods, decreasing unnecessary alerts that could interrupt clinical work [9]. Workflow interruption analysis documented a significant reduction in security-related workflow disruptions following implementation, from multiple interruptions per day to far fewer across clinical departments. The average duration of necessary interruptions also decreased notably, reflecting more efficient security protocols that minimized clinical impact. Staff satisfaction surveys revealed substantial improvement in satisfaction with security measures after implementation of the context-aware framework [7]. This improvement was primarily attributed to the reduction in disruptive false alarms and unnecessary interventions that had previously impeded clinical workflows. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1181 Response time metrics showed the average duration from threat detection to resolution decreased considerably, a substantial improvement attributable to the framework's automated response orchestration capabilities [8]. This rapid response capability proved particularly valuable for containing potential security incidents before they could impact clinical operations. The framework's graduated response mechanisms implemented proportional security measures based on threat severity and clinical context, avoiding unnecessary disruption of critical care activities. This contextsensitive approach proved especially beneficial in high-acuity environments such as emergency departments and surgical units. The framework's ability to maintain essential functions during security incidents represented a significant advancement over traditional approaches that often implement binary blocking decisions. During simulated security incidents, the system maintained the vast majority of critical functionality through automated workflow reconfiguration and resource reallocation. This resilience-focused design reflected the recognition that in healthcare environments, availability often takes precedence over other security considerations due to potential patient safety implications of service disruptions. 3.5. Technical Performance Results The framework demonstrated reasonable resource requirements and favorable scalability characteristics across all implementation sites. Computational overhead increased system resource utilization moderately compared to baseline operations, representing an acceptable impact given the substantial security improvement [8]. This modest overhead resulted from the framework's efficient implementation of detection algorithms and strategic distribution of processing loads across system components. The distributed architecture effectively leveraged existing infrastructure capacities, minimizing additional hardware requirements for most implementation sites. Scalability analysis revealed predominantly linear scaling of resource requirements relative to institution size and transaction volume, enabling accurate capacity planning for future deployments. Integration effort averaged substantial person-hours per facility, with significant variation based on the complexity of existing systems and the extent of custom integration requirements [5]. Facilities with standardized, modern infrastructure completed implementation with considerably less effort than those with higher proportions of legacy systems. Knowledge transfer from earlier implementations significantly reduced effort requirements for subsequent deployments, demonstrating valuable organizational learning effects. Maintenance requirements stabilized at reasonable levels for ongoing monitoring and tuning, with larger facilities generally requiring more maintenance attention. The framework's continuous learning capabilities reduced manual maintenance requirements over time, with alert tuning needs decreasing steadily as the system adapted to each institution's specific operational patterns [9]. This adaptive learning capability proved particularly valuable for accommodating the frequent changes in clinical workflows characteristic of healthcare environments. Performance evaluation demonstrated that AI-based security systems outperformed traditional approaches across key metrics including detection accuracy, false positive rates, and response times [8]. The framework's context-aware security approach proved particularly valuable in healthcare environments characterized by complex, variable workflows and high consequences for security failures. The evaluation results validated the core design principles underlying the framework and indicated substantial advantages over traditional security approaches for protecting AIenabled healthcare systems. 4. Conclusion The AI-Augmented Workflow Resilience Framework addresses the distinctive cybersecurity challenges facing healthcare organizations as they increasingly adopt artificial intelligence technologies. By embedding security mechanisms directly into clinical and administrative workflows, the framework successfully balances robust protection with operational efficiency—a critical consideration in healthcare environments where system availability directly impacts patient care. The five-layer architecture provides comprehensive coverage against both traditional and AIspecific threats while maintaining sensitivity to the unique operational demands of healthcare settings. Implementation across multiple healthcare environments demonstrated the framework's adaptability to diverse institutional contexts and technological infrastructures. The context-aware approach proved particularly valuable in reducing false positives and unnecessary workflow disruptions, addressing common limitations of conventional security systems that lack healthcare-specific knowledge. The framework's graduated response mechanisms and workflow reconfiguration capabilities ensured the preservation of essential clinical functions during security incidents, reflecting the priority of service availability in healthcare environments. World Journal of Advanced Research and Reviews, 2025, 26(02), 1175-1182 1182 The evaluation results validate the framework's core design principles and confirm substantial advantages over traditional security approaches for protecting AI-enabled healthcare systems. The demonstrated improvements in threat detection, response orchestration, and operational impact address a critical gap in current healthcare cybersecurity practices. Furthermore, the framework's continuous learning capabilities enable ongoing adaptation to emerging threats and evolving clinical workflows, ensuring sustained effectiveness in dynamic healthcare environments. This work contributes to the advancement of healthcare cybersecurity by providing a specialized approach that acknowledges and addresses the unique challenges of securing AI systems in clinical settings. By protecting these technologies while supporting rather than impeding their beneficial applications, the framework enables healthcare organizations to realize the benefits of AI innovation while mitigating the associated security risks. Future developments will focus on extending the framework to encompass additional AI applications in healthcare and further optimizing the balance between comprehensive security monitoring and efficient clinical operations. References [1] Anthony Musk, Harold Castro, “Cybersecurity Challenges in AI-Enabled Healthcare Systems,” January 2025, Research Gate, Available: https://www.researchgate.net/publication/389500963_Cybersecurity_Challenges_in_AIEnabled_Healthcare_Systems [2] Wendy Burke, et al, “From Dis-empowerment to empowerment: Crafting a healthcare cybersecurity selfassessment,” Computers & Security, Volume 148, January 2025, Available: https://www.sciencedirect.com/science/article/pii/S016740482400453X [3] Gerda Bortsova, et al, “Adversarial attack vulnerability of medical image analysis systems: Unexplored factors,” Medical Image Analysis, Volume 73, October 2021, Available: https://www.sciencedirect.com/science/article/pii/S1361841521001870 [4] Julia Stefanie Roppelt, et al, “Artificial intelligence in healthcare institutions: A systematic literature review on influencing factors,” Technology in Society, Volume 76, March 2024, Available: https://www.sciencedirect.com/science/article/pii/S0160791X23002488 [5] Reshma Vemula, “Transforming Healthcare it: A Technical Analysis of AI Integration and Implementation,” February 2025, INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING & TECHNOLOGY, Available: https://www.researchgate.net/publication/388875184_Transforming_Healthcare_it_A_Technical_Analysis_of_ AI_Integration_and_Implementation [6] Paul Koster, “Security Analytics and Monitoring of Medical Devices,” now publishers, 15 Sep 2021, Available: https://nowpublishers.com/article/Chapter/9781680838220?cId=978-1-68083-823-7.ch17 [7] Chandra Sekhar Veluru, “Impact of Artificial Intelligence and Generative AI on Healthcare: Security, Privacy Concerns and Mitigations,” February 2024, Journal of Artificial Intelligence & Cloud Computing, Available: https://www.researchgate.net/publication/382193776_Impact_of_Artificial_Intelligence_and_Generative_AI_o n_Healthcare_Security_Privacy_Concerns_and_Mitigations [8] Hind Meziane, Noura Ouerdi, “A survey on performance evaluation of artificial intelligence algorithms for improving IoT security systems,” December 2023, Research Gate, Available: https://www.researchgate.net/publication/376153572_A_survey_on_performance_evaluation_of_artificial_inte lligence_algorithms_for_improving_IoT_security_systems [9] Emma Oye, Paul Lois, “The Role of Artificial Intelligence in ContextAware Healthcare: Enhancing DecisionMaking and Predictive Analytics,” September 2023, Research Gate, Available: https://www.researchgate.net/publication/389052518_The_Role_of_Artificial_Intelligence_in_Context_Aware_Healthcare_Enhancing_Decision-Making_and_Predictive_Analytics