International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.

1, March 2025

SECURING TELEHEALTH PLATFORMS:

ML-POWERED PHISHING DETECTION WITH
DEVOPS IN HEALTHCARE ANALYTICS
Jesu Marcus Immanuvel Arockiasamy

Engineer Lead Sr. & DevOps Expert, Healthcare Analytics, Leading

Healthcare Company, Richmond, Virginia

ABSTRACT
The rapid growth of telehealth due to recent global health crises has highlighted its
vulnerabilities, particularly to phishing attacks. These attacks exploit both technical and
user behaviour gaps and pose significant risks to sensitive health information. We
demonstrate a hybrid ML model (Random Forest + XGBoost) that detects phishing URLs
in telehealth portals with 93% accuracy, validated through a browser plugin in real time.
We integrate this model into a telehealth-optimized DevSecOps pipeline, enhancing
security measures. A notable case study showcases a browser plugin ("Phish & Chips")
that blocks malicious portals pre-login while automating compliance audits. Our approach
employs real-time data, behavioural analytics, and EHR integration to harden defences,
improve detection, accelerate response times, and ensure HIPAA compliance. This
approach is not only good for security and operational resilience but also a scalable and
cost-effective solution for telehealth platforms.

KEYWORDS
Phishing, Machine Learning, DevOps, Telehealth, Healthcare Analytics

1. INTRODUCTION: THE TELEHEALTH SECURITY PARADOX

1.1. Overview of Telehealth Expansion and Increased Cybersecurity Threats

Telehealth has transformed traditional healthcare; 87% of US hospitals now offer virtual care [1].
This rapid growth - fueled by the COVID-19 pandemic—has given patients access to remote
consultations, prescription management, and real-time health monitoring. However, this digital
transformation has also exposed healthcare to unprecedented cyber threats. Post 2020, healthcare
phishing attacks skyrocketed by 189% (CISA, 2023) [2], and criminals are exploiting the sector's
non-tech savvy users (e.g., elderly patients) and high-value data. Traditional security measures
like blocklists and signature-based detection fail against AI-generated spear-phishing attacks on
prescription portals and patient portals. The healthcare industry faces a paradox: patient
accessibility vs robust security.

1.2. Why Phishing is a Big Threat to Healthcare Platforms

Phishing has become the preferred method for hackers to get into healthcare organizations to
steal medical data and/or deploy ransomware, with 40% of overall breaches (APWG, 2023) [3].
These attacks often come disguised as urgent medical alerts, pharmacy refills, or vaccine

DOI: 10.5121/ijbb.2025.15103 21
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
schedules and aim to trick staff and patients into sharing personal credentials, stealing medical
data, or downloading malicious software. For example, phishing emails with ransomware forced
hospitals to return to paper records, delaying treatment and costing $10.93 million per breach
(IBM Security, 2023) [4]. Verizon's 2023 Data Breach Investigations Report shows 93% of
socially engineered attacks in healthcare are phishing. The industry's unique challenges – high
data sensitivity, low user security literacy, and regulatory pressure – make it a prime target for
cyberattacks [5].

Figure 1 - Common Phishing Scenarios in Healthcare

Given the escalating threat of phishing attacks and the limitations of traditional security
measures, a proactive and adaptive approach is essential. To address these challenges, we
propose an innovative security framework that integrates Machine Learning (ML)-driven
phishing detection into DevOps pipelines. By leveraging real-time analytics and behavioural
analysis, our solution aims to enhance the security posture of telehealth platforms while
maintaining patient accessibility.

1.3. DevOps and Securing Telehealth Apps

To stay ahead of the threats, telehealth platforms need adaptable security. DevOps with
DevSecOps puts security into the software development lifecycle. CI/CD pipelines enable real-
time security updates, while automated testing tools like SAST and DAST find vulnerabilities
during pre-deployment. Integrating phishing detection into CI/CD workflows allows rapid
patching of zero-day exploits. DevOps fosters collaboration between development and security
teams, which accelerates the implementation of countermeasures—for example, promptly
enforcing multi-factor authentication when signs of credential harvesting are detected.
DevSecOps reduces mean time to recovery (MTTR) by 63% (Gartner, 2023) [6], critical for the
uptime of life-critical telehealth services.

1.4. Ml-Driven Phishing Detection

Machine learning (ML) introduces a paradigm shift in phishing defence, by replacing static rule-
based methods with dynamic, behaviour-based analysis. ML models trained on telehealth-
specific data—user interaction patterns, email metadata, URL structures—detect anomalies that

22
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
indicate phishing. For example, NLP algorithms flag emails that mimic clinical urgency but have
malicious payloads. Real-time classifiers are 98% accurate at detecting fraudulent login pages
(IEEE, 2023) [7] and block access before credentials are compromised. When integrated with
DevOps pipelines, these models self-update through feedback loops and adapt to new attack
vectors. A behavioural-first approach also reduces false positives by looking at contextual clues
(e.g., a patient's typical login time) and minimizing disruption to care.

Synthesis of Key Points

- Background: Telehealth has grown faster than legacy security has kept up, leaving gaps to
be exploited.

- Problem Statement: Phishing exploits healthcare's social and technical vulnerabilities and
requires a dynamic solution.

- Literature: Studies have shown blocklists do not work (IBM, APWG), and automation is
the way to go (Verizon, Gartner).

- Solution: DevOps-ML is the continuous, adaptive phishing defense.

- Innovation: Combining behavioral ML and DevSecOps creates a self-learning security

layer that reduces reliance on human vigilance and static rules.

This white paper will introduce a new framework for operationalizing ML-powered phishing
detection in DevOps. It will give healthcare providers a scalable defence against the $10M-a-
breach while keeping telehealth accessible.

2. LITERATURE SURVEY
Phishing detection remains a cat-and-mouse game in cybersecurity, with attackers constantly
evolving their tactics to bypass traditional defences. Phishing detection techniques are
categorized into three main approaches: blocklist-based, heuristic-based, and AI/ML-based
detection.

2.1. Phishing Detection Techniques

Blocklist-Based Detection:

This method uses a list of known phishing URLs to block malicious sites. When a user tries to
access a URL, the system checks it against a predefined list and blocks it if it is marked as
phishing. Although this is simple and works for known threats, it is reactive and does not handle
new phishing URLs.

Heuristic-Based Detection:

Heuristic-based detection uses predefined rules to identify potential phishing sites. These rules
analyse elements such as domain age, presence of suspicious characters, and deviation from
standard webpage structure. This approach is proactive but has high false positive rates (false
positives - flagging legitimate sites as phishing) and requires continuous updates [8].

23
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
AI/ML-Based Detection:

AI and machine learning (ML) approaches detect phishing attempts by analysing large datasets
and identifying patterns. These methods adapt to new threats and are generally accurate.
However, they require substantial computational resources and high-quality training data. This
approach uses machine learning models to detect phishing by analysing large datasets. It is
adaptive and highly accurate but dependent on resources and high-quality data.

Table 1 - Phishing Detection Techniques – Comparison

Detection
Advantages Disadvantages
Approach
- Reactive; easily bypassed with new
- Simple and effective for known URLs.
Blocklist-based
threats. - Lag in updates can create vulnerability
gaps.
- Proactive and customizable - High false-positive rate.
Heuristic-based
detection. - Requires continuous rule management.
- Adaptive learning and high - Resource-intensive and data-dependent.
AI/ML-based
accuracy.

Hybrid approaches combining these methods show promise but are rarely integrated into
DevOps workflows [9] for telehealth platforms —a gap this paper addresses. Following diagram
illustrates the hybrid approach which combines these three approaches for phishing detection:

Figure 2 - Proposed Hybrid Phishing Detection Approach

24
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
2.2. URL Feature Extraction: Balancing Accuracy and Efficiency

Feature extraction plays a pivotal role in phishing detection by identifying the most critical
attributes distinguishing phishing URLs from legitimate ones. The more relevant features
included in model training, the better the model's accuracy in detecting phishing attempts. While
prior work leverages 30+ features (e.g., domain age, SSL validity), our analysis of the PhiUSIIL
dataset (134,850 legitimate vs. 100,945 phishing URLs [10]) revealed that 8 key features achieve
99% of the discriminative power of full models (Table 2). This enables lightweight deployments
like a simple browser plugin without sacrificing efficacy.

Features Used in this PoC:

For the proof-of-concept (PoC), we extracted and used the following eight features. (Beyond the
PoC, a more extensive set of features that can enhance phishing detection models are included in
Appendix section)

Table 3 - Features Used in this PoC

Feature Description Rule

If the domain part contains an IP Domain name contains IP → Phishing
IP Address
address, it is likely phishing
Longer URLs are often associated Length < 75 → Legitimate; Length ≥ 75 →
URL Length
with phishing Phishing
URL Shortened URLs may conceal Tiny URL-like service → Phishing
Shortening malicious intent
Special If the URL contains special Multiple Special Characters → Phishing
Characters characters such as @,"", /, _, -
Subdomain Excessive subdomains may indicate 1 dot → Legitimate; 2 dots → Suspicious; 3+ dots
Count phishing → Phishing
HTTPS + Trusted Issuer + Certificate Age ≥ 1
SSL State Valid SSL certificates improve trust
Year → Legitimate; Otherwise → Phishing
Google-indexed sites are generally Indexed by Google → Legitimate
Google Index
legitimate
Website Higher-ranked sites are less likely to Rank < 100,000 → Legitimate; Rank > 100,000
Traffic be phishing → Suspicious; Otherwise → Phishing

By extracting and analysing these features, machine learning models can detect phishing URLs
with improved precision.

2.3. ML Algorithms Comparison

For this study, we trained multiple ML models on the PhiUSIIL Phishing URL Dataset, which
consists of 134,850 legitimate URLs and 100,945 phishing URLs. Ten different algorithms were
evaluated for accuracy and performance: (emphasis was more on measuring metrics critical for
healthcare portals: Recall (to avoid missed threats) and FPR (to prevent care disruptions).

25
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
Table 3 - ML Algorithms Comparison

Model Accuracy Precision Recall F1 Score MCC FPR FNR

Random Forest 99.44 99.26 99.76 99.51 0.99 0.01 0.00
Decision Tree 99.39 99.22 99.7 99.46 0.99 0.01 0.00
KNN 99.14 98.97 99.52 99.25 0.98 0.01 0.00
Logistic Regression 98.72 98.37 99.4 98.88 0.97 0.02 0.01
Naive Bayes 93.54 91.61 97.6 94.51 0.87 0.12 0.02
AdaBoost 99.10 98.92 99.51 99.21 0.98 0.01 0.00
Gradient Boosting 99.24 98.97 99.69 99.33 0.98 0.01 0.00
XGBoost 99.44 99.19 99.84 99.51 0.99 0.01 0.00
LightGBM 99.43 99.15 99.84 99.5 0.99 0.01 0.00
Cat Boost 99.39 99.09 99.84 99.46 0.99 0.01 0.00

From this comparison, Random Forest and XGBoost demonstrated the highest accuracy. To
further optimize the model, we combined the strengths of both models using a stacking ensemble
to achieve improved accuracy [11]. The scripts used for model comparison are included in the
GitHub repository, as referenced in the Appendix section.

2.4. Integrating Phishing Detection into Devops Pipelines

To enhance security automation, this phishing detection ML model can be integrated into
DevOps workflows:

1. Data Pipeline: Continuously fetch, clean, and preprocess URL data for real-time model
updates.

2. Model Training & Evaluation: Automate model retraining using MLOps frameworks
(e.g., Kubeflow, MLflow).

3. Containerization: Deploy trained models in Docker containers for consistent runtime

environments.

4. CI/CD Integration: Incorporate phishing detection into CI/CD pipelines, scanning

URLs before deploying in any environment.

5. Browser Plugin Deployment: Automate deployment of browser extensions that use the
ML model to warn users before accessing suspicious sites.

6. Monitoring & Feedback Loop: Implement logging and user feedback mechanisms to
improve model accuracy over time.

3. METHODOLOGY: DEVOPS-DRIVEN PHISHING DEFENCE

3.1. Architecture Overview

Our framework integrates ML phishing detection into telehealth DevOps pipelines through three
components:

26
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
1. Automated Model Training:

o Process – Daily training on phishing datasets (like Phishtank or OpenPhish

databases)

o Tools - Docker containers for reproducibility and MLflow for tracking metrics.

2. Real-Time Browser Plugin

o Deployment – Random Forest + XGBoost model (with at least 8 features) packaged

as a Chrome extension.

o Function - Blocks suspicious telehealth portals pre-login

3. DevOps Feedback Loop

o Incident Response - Detected threats trigger Jira tickets for security teams

o CI/CD Integration - AWS Lambda updates blocklists and heuristic rules nightly,
OpenShift deployments for API hosting

3.2. Case Study Preview: Browser Plugin in Action

To validate the effectiveness of ML-powered phishing detection, we developed Phish & Chips, a
browser-based phishing prevention system as part of the MLX Hackathon 2025 [12]. This real-
world proof-of-concept integrates an AI-powered browser plugin with a Flask-based ML model,
providing real-time phishing alerts for healthcare platforms.

Plugin Architecture & Workflow:

Figure 4 - Phish & Chips Browser Plugin Architecture & Workflow

27
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
The Phish & Chips browser extension offers real-time protection by detecting and alerting
users to phishing attempts. It functions by intercepting the URL of a website visit, extracting key
phishing indicators, and using a Random Forest model or a stacked model via a Flask API to
classify the URL as Legitimate or Phishing. Users receive visual alerts with icons indicating site
safety: a green shield for safe sites, a red warning for suspicious ones, and a redirection to a
warning page.

Dataset Used:

The ML model was trained on a dataset of 250K URLs from the UCI Machine Learning
Repository, enhanced with additional trusted domains from Similar Web (refer Appendix
section for the dataset details).

Tools Used:

- Machine Learning: Scikit-learn (Random Forest, XGBoost), NumPy, Pandas

- API Backend: Flask for inference, FastAPI (potential optimization)
- Browser Extension: JavaScript, Manifest V2 (for Firefox compatibility)
- DevOps Pipeline: GitHub Actions (CI/CD for model updates), PyTest (unit testing),
OpenShift (hosting for scalability)

Feature Importance Analysis:

Feature importance is a key aspect in machine learning that identifies which variables or features
in a dataset are most influential in making predictions. By assigning an importance score to each
feature, it helps in understanding the weight and significance of each feature on the model's
output. This can aid in refining the model by focusing on the most impactful features, ultimately
enhancing the model's accuracy and interpretability.

Table 4 - Trained ML Model Feature Importance Score

Is Subdomain URL Special Domain Is

Feature
HTTPS Count Length Characters Reputation Trusted
Importance
0.4 0.27 0.16 0.07 0.05 0.02
Score

To address the low trust score of the "Is Trusted" feature for trusted sites, it's clear that
adjustments are needed for better accuracy. You can modify this feature programmatically by
adjusting the feature importance in your model. Emphasizing "Domain Reputation" and "Is
HTTPS" by raising their weights could improve the identification of legitimate sites. This
adjustment will help the model prioritize trust indicators more effectively, leading to better
classifications [13]. Additionally, the Flask API allows other fields to be modified in importance
as needed to fine-tune the model's performance.

Results & Impacts:

 Real-time phishing detection achieved 93% accuracy.

 Browser extension successfully prevented phishing attacks in test scenarios.
 Demonstrated seamless integration of ML and DevOps into a telehealth security pipeline.

With the success of Phish & Chips, the next step is integrating phishing detection into DevOps
workflows, ensuring security is automated in telehealth applications.

28
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025

4. DEVOPS-INTEGRATED PHISHING DETECTION: A SECURE PIPELINE

APPROACH
4.1. Embedding ML-Based Phishing Detection into DevOps CI/CD Pipelines

Traditional security gates in CI/CD (e.g., SAST/DAST) lack context for healthcare-specific
phishing patterns, such as fake patient portals or fraudulent prescription refill links. The solution
for that problem is a telehealth-optimized DevSecOps pipeline using the Phish & Chips model,

Table 5 - Telehealth Optimized DevSecOps Pipeline

Feature Action Tool Outcome/Impact

Utilize the Random
GitHub Actions with Successfully blocked a phishing
Pre-Commit Forest model to scan
a custom Python URL during a code review,
Security Scans URLs in code during
script. preventing potential breaches.
pull requests.
Post-Deployment Real-time tracking of OpenShift-hosted Detected 12 phishing attempts
Threat phishing attempts via Flask API with within 48 hours, triggering
Monitoring browser plugins. Prometheus alerts. immediate security responses.
Weekly retraining of MLflow and AWS
Automated Improved model accuracy by 8%
models with new Lambda for
Model Retraining over
phishing patterns. serverless retraining.

4.2. Compliance Automation for Healthcare

In healthcare, HIPAA requires audit trails for security-related incidents but manually logging
them is error-prone and time-consuming. This approach effectively handles compliance-related
issues.

 Automated Audit Reports: Generate compliance documentation directly from browser

plugin logs.
 Breach Documents: Use Phish & Chips triggered Jira Tickets to auto-populate incident
details (like IP addresses, URLs, and timestamps)

4.3. How Phishing Detection Fits into CI/CD Pipelines

Adding phishing detection to CI/CD pipelines is crucial for keeping telehealth platforms secure.
It allows us to catch threats early, reducing the risk of attacks. Following Table 6 shows how we
integrate this process into DevOps to manage risks and automate responses.

Table 6 - Phishing Detection in DevOps Processes

DevOps Component Phishing Detection Integration

CI/CD Pipelines Auto-scan URLs in web apps before deployment
Automated Testing SAST/DAST scan API responses for phishing
Monitoring & Alerts AI-based phishing analytics in production
Incident Response Security playbooks trigger actions on threats

- Continuous Threat Updates – The ML model fetches live phishing feeds (Google Safe
Browsing, PhishTank) for auto-updates.

29
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
- Security-First CI/CD Hooks - Before deployment, the system scans embedded links in
telehealth portals to detect potential phishing.
- Automated Risk Classification - If phishing threats are detected, CI/CD stops the
release, flagging it for security review.

4.4. Automating Security Testing in DevSecOps and Continuous Monitoring

Adding automated security checks and ongoing monitoring to our DevSecOps practices helps us
spot and tackle cybersecurity threats before they become issues. By using smart technologies like
machine learning and advanced monitoring tools, telehealth platforms are always ready to fend
off phishing attacks and other security risks. The table below breaks down these efforts, showing
how each part plays a role in keeping us safe.

Table 7 - Key Components of Automated Security Testing and Monitoring

Component Description
Unit Tests for Safe URL
CI/CD Validates that all links in telehealth apps match trusted site criteria
Patterns
The ML model scans API responses for phishing redirects or fake login
API Security Testing
attempts
Regression Testing for
The CI/CD systems retain models when new phishing attacks are detected
New Threats
AI Driven Monitoring Track phishing attempts across telehealth systems, triggering responses like
Dashboards access restrictions and MFA.
Automated Incident
Block malicious domains instantly and report it as an incident with details
Creation
Distribute phishing alerts across healthcare networks, this approach ensures
Threat Intelligence Sharing
dynamic and robust security posture.

5. REAL-TIME HEALTHCARE ANALYTICS FOR PHISHING PREVENTION

Phishing attacks on telehealth systems need quick action. Adding real-time analytics not only
helps spot these attacks faster but also reduces their impact. This section will explain how using
live data, behavioural insights, and connecting with electronic health records can fight phishing
and it will also offer ideas for making things even better.

5.1. Real-Time Data Streaming and Analysis

Key Components:

Table 8 - Real-Time Data Streaming - Key Components

Component Function Tools/Techniques

Kafka-Based Ingest live telehealth session data for
Apache Kafka
Streaming continuous analysis.
Custom stacked ML models,
Dynamically scan URLs and data streams to
AI Engine Random Forest, real-time
detect phishing patterns in real time.
inference
Analyse user behaviour (e.g., login times,
Telehealth
geolocation mismatches) to flag suspicious Elasticsearch, Kibana dashboards
Monitoring
activity immediately.

30
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
Implementation Highlights:

- Live Ingestion: Kafka streams continuously capture telehealth session data [14].
- Real-Time Scanning: The AI engine examines URLs and session patterns, issuing instant
warnings when suspicious activity is detected.
- Behavioral Analytics: Abnormal login behaviors (e.g., off-hour access, geolocation
anomalies) trigger further scrutiny and automated alerts.

5.2. Anomaly Detection and EHR Integration

Detection Approaches:

 Behavioral Analytics in Patient Portals:

o Monitor login timestamps and geolocation data to flag anomalous access (e.g., an
elderly patient logging in at an unusual hour).
o Case Example: Correlating geolocation data with prescription refill requests has
previously blocked multiple fraudulent attempts.

 Predictive Threat Intelligence:

o Cluster phishing attempts by campaign type (e.g., COVID-19 test scams) and
proactively update security measures.
o Utilize time-series forecasting (e.g., using Prophet) to predict attack spikes and
adjust defenses preemptively.
o Outcome: Such measures have reduced phishing success rates significantly
during peak threat periods.

 EHR Data Correlation:

o Cross-reference login IPs with patient records to identify imposters.

o For instance, if a login originates from a high-risk node (e.g., a Tor exit point),
the system triggers multi-factor authentication (MFA) and alerts.
o Tool Integration: FastAPI middleware acts as a bridge between EHR systems
(e.g., Epic, Cerner) and the phishing detection ML Model.

5.3. Future Directions and Impact

We have several areas to focus on moving forward:

 Sharper Machine Learning: To increase detection, refine ML models with evolving

phishing tactics, and add more data sources (e.g., social media signals).
 Smarter Incident Handling: Create more advanced security playbooks that alert and
autonomously contain threats by isolating sessions or accounts.
 Threat Intelligence Sharing: Collaborate with other healthcare networks to share
anonymized threat data to increase sector resilience.
 Scalability and Performance: Invest in scalable architectures to handle growing data for
low-latency detection and response as telehealth grows.

31
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
Impact Recap:

 Better Detection: Real-time analytics catches phishing attempts that static models usually
overlook.
 Quicker Reactions: Automated, immediate action reduces the window of vulnerability.
 Less Damage: Proactive and predictive measures reduce financial and reputational
damage from phishing.

Integrating real-time analytics into our security framework enables swift detection and mitigation
of phishing threats. This integrated approach not only hardens the defences now but also enables
continuous improvement and collaboration to protect your sensitive health data.

6. CONCLUSIONS
Our strategy strengthens the security of telehealth platforms by emphasizing real-time analysis
and adaptive learning. The development of the Phish & Chips browser plugin shows the practical
application of machine learning to mitigate phishing in healthcare. Integrating ML-based security
tools into the DevOps pipeline improves early detection, prevents cyber-attacks, and ensures
continuous compliance. This automation and feedback mechanism creates a resilient security
framework that protects patient data and ensures efficient telehealth services.

Going forward, we will refine ML models, add more data sources and collaborate with healthcare
networks to share threat intelligence. Our goal is to create a dynamic security ecosystem that not
only addresses current threats but also pre-empts future cyber risks like Mishing (attacks that use
SMS instead of emails) [15]. These advanced security measures are critical to maintaining the
integrity and reliability of telehealth services as a trusted resource for patients and healthcare
providers.

In conclusion, embedding ML-based phishing detection within DevOps workflows represents a

transformative advancement for securing telehealth platforms. Our framework uses real-time
analytics, behavioural anomaly detection, and EHR data correlation to defend against evolving
threats comprehensively. As threats grow, we will enhance our models, automate incident
response, and share threat intelligence to stay resilient.

REFERENCES

[1] American Hospital Association (AHA) publication, Fact Sheet: Telehealth, Issue - February 2025.
https://www.aha.org/fact-sheets/2025-02-07-fact-sheet-telehealth
[2] Cybersecurity & Infrastructure Security Agency (CISA) (2023). Cybersecurity Threats in Healthcare.
Retrieved from https://www.cisa.gov/healthcare
[3] Anti-Phishing Working Group (APWG) - Phishing Activity Trends Report 2023 – Reported by Anti-
Phishing Working Group. Retrieved from https://www.apwg.org/report
[4] Cost of a Data Breach Report 2023 by IBM Security. Retrieved from
https://www.ibm.com/reports/data-breach
[5] 2023 Data Breach Investigations Report (DBIR) by Verizon Network Center – Available at
https://www.verizon.com/business/resources/reports/2024-dbir-data-breach-investigations-report.pdf
[6] Gartner.com (2023). DevSecOps Adoption in Healthcare: Trends and Best Practices
[7] H. Sharma, P. Sharma and R. Singh, "Real-Time Phishing Attack Detection through Advanced
Machine Learning Techniques," 2023 IEEE International Conference on ICT in Business Industry &
Government (ICTBIG), Indore, India, 2023, pp. 16, DOI:10.1109/ICTBIG59752.2023.10456013
[8] Rajaram, S. K., Konkimalla, S., Sarisa, M., Gollangi, H. K., Madhavaram, C. R., Reddy, M. S.,
(2023). AI/ML-Powered Phishing Detection: Building an Impenetrable Email Security System. ISAR
Journal of Science and Technology, 1(2), 10-19.
32
 International Journal on Bioinformatics & Biosciences (IJBB) Vol 15, No.1, March 2025
[9] Arockiasamy J.M.I. (2025) Proactive Healthcare Analytics: Early Detection of Diabetes with SDOH
Insights and Machine Learning, European Journal of Computer Science and Information Technology,
13 (2), 64-74
[10] Prasad, A. & Chandra, S. (2024). PhiUSIIL Phishing URL (Website) [Dataset]. UCI Machine
Learning Repository. https://doi.org/10.1016/j.cose.2023.103545.
[11] Kavakiotis, I., Tsave, O., Salifoglou, A., Maglaveras, N., Vlahavas, I., & Chouvarda, I. (2017).
Machine Learning and Data Mining Methods in Diabetes Research. Computational and Structural
Biotechnology Journal, 15, 104-116.
[12] MLX (Machine Learning Xtreme) Hackathon - https://mlx-hack-2025.devpost.com
[13] Najla Odeh, Derar Eleyan, Amna Eleyan, "Enhancing Web Security through Machine Learning-based
Detection of Phishing Websites", International Journal of Computer Network and Information
Security (IJCNIS), Vol.17, No.1, pp.39- 56, 2025. DOI:10.5815/ijcnis.2025.01.04
[14] Arockiasamy, Jesu Marcus Immanuvel. (2025). DevOps-Driven Real-Time Health Analytics: A
Scalable Framework for Wearable IoT Data. International Journal for Multidisciplinary Research.
Volume 7. 10.36948/ijfmr.2025.v07i01.37358.
[15] Jordyn Alger, Cybersecurity - Mobile phishing threats – Retrieved from
https://www.securitymagazine.com/

AUTHOR

Jesu Marcus Immanuvel Arockiasamy is a seasoned Healthcare Analytics and

DevOps expert with over 18 years of experience at a leading healthcare company. His
work focuses on leveraging DevOps principles to enhance system efficiencies,
automate deployments, and manage CI/CD pipelines with prominent tools like
Jenkins, Kubernetes, Terraform, and AWS. A devoted mentor, he has fostered a
collaborative DevOps culture that promotes innovation and agility.

Arockiasamy has published several impactful whitepapers, including "Digital Healthcare Evolution: The
Power of DevOps for Better Patient Engagement" and "Proactive Healthcare Analytics: Early Detection of
Diabetes with SDOH Insights and Machine Learning." These works explore integrating advanced analytics
and machine learning into healthcare solutions, aiming to enhance patient care and engagement. His
academic and professional endeavors continue to drive transformative strategies in healthcare technology,
ensuring secure, scalable, and patient-centric digital solutions.

APPENDIX

 GitHub repository with Working Code: Access the repository to explore the codebase and practical
implementations discussed throughout this document.
 Culinary-Inspired Hackathon demo video of Phish & Chips Browser Plugin: Watch this demo
video to see the creative presentation of the browser plugin functionality in action, inspired by
culinary themes.
 All 30 Features Extracted from URL Validation: Review a comprehensive list of features used for
URL validation, offering technical insights into our phishing detection criteria.

Datasets Used: PhiUSIIL Phishing URL SimilarWeb Top 500 Websites

33