2025_EML Capstone Projects

Comparison between frameworks: Framework Disclosure

Constraints

SWOT

This framework displayed a strong alignment with VA’s long-term organizational objectives. It showed each solution’s proper fit in detail. Internal and external factors are identified for the best strategic solution Extensive review of compliance, operational obstacles and feasibility analysis.

Potential false positives might create alert fatigue and an extra burden for the security team. Stakeholders’ trust in AI can be considered as constraint

Risk Assessment

Infrastructure costs, risk with project timeline and regulatory scrutiny.

Two selected frameworks provided clear visibility and realistic vision to execute the project.

Selected Solution: Based on analysis, Solution #1, machine learning-based anomaly detection would be recommended to implement for the project. It aligns with my project overall objectives and goals as it intends to build innovative cybersecurity systems with the VA organization. This approach also supports VA’s future AI driven technological capabilities by preparing the organization. Solution #2, rule-based scoring would be recommended as a secondary option when project timelines are tight, and resources are limited. Shifting to this solution would be ideal if constraints occur to complete the implementation faster as it can be governed easier. 7. Risk and Change Management Risk Management Plan: The project has high sensitivity due to cybersecurity, personal data privacy and organizational trust, therefore, a comprehensive risk management plan was a necessity to ensure robust development and deployment processes. Operational risk: False positives cause significant amounts of workflow disruption and alert fatigue and delays in mitigating critical threats. Mitigation strategy: It includes incorporating anomaly severity scoring to prioritize investigations and tuning detection thresholds based on feedback from testing processes. Another strategy is to build machine learning human hybrid loops as low-confidence detections are routed for manual reviews.

Back to Table of Contents