SHIELD SMART: Explainable Phishing Detection for IoT

Inspiration

As IoT devices become increasingly integrated into homes, industries, and critical infrastructure, they have also become attractive targets for cybercriminals. Phishing attacks remain one of the most common attack vectors, often leading to credential theft, malware infections, and unauthorized access. We were inspired to develop SHIELD SMART to address the growing need for intelligent and explainable phishing detection solutions that can protect both digital and physical IoT environments.

What It Does

SHIELD SMART is an AI-powered cybersecurity framework that detects phishing attacks in IoT ecosystems through a multi-layered approach. The system analyzes suspicious URLs and text content using deep learning models to identify malicious patterns. In addition to detecting threats, the framework provides explainable insights that help users understand why a URL or message has been classified as phishing. The platform can also trigger real-time security responses through IoT-enabled devices, improving overall threat awareness and protection.

How We Built It

The project combines multiple technologies and AI models:

CNN-LSTM model for phishing URL analysis and classification.
DistilBERT model for semantic understanding of suspicious text content.
Explainable AI (XAI) techniques to generate human-readable explanations for model predictions.
Python for model development and integration.
Machine learning and deep learning frameworks for training and evaluation.
ESP32-based IoT integration for real-time alert generation and physical response mechanisms.
Web dashboard for monitoring detected threats and viewing explanations.

Challenges We Ran Into

Building an effective phishing detection system presented several challenges:

Collecting and preparing high-quality phishing and legitimate datasets.
Handling evolving phishing techniques that continuously change attack patterns.
Balancing model accuracy with computational efficiency for real-time detection.
Making deep learning predictions interpretable through Explainable AI methods.
Integrating software-based threat detection with IoT hardware components.

Accomplishments That We're Proud Of

Successfully developed a hybrid AI framework for phishing detection.
Integrated Explainable AI to improve transparency and user trust.
Achieved strong detection performance using deep learning techniques.
Designed a scalable architecture suitable for future IoT security applications.
Connected AI-driven threat detection with real-world IoT response mechanisms.

What We Learned

Through this project, we gained valuable experience in:

Deep learning model development and optimization.
Cybersecurity threat detection techniques.
Explainable Artificial Intelligence (XAI).
IoT security architecture and device integration.
Building practical AI solutions that address real-world security challenges.

What's Next for SHIELD SMART

Future enhancements include:

Real-time threat intelligence integration.
Support for additional phishing attack vectors.
Mobile application development.
Cloud-based deployment for large-scale IoT environments.
Advanced behavioral analytics for proactive threat prevention.

SHIELD SMART demonstrates how AI, cybersecurity, and IoT technologies can work together to create smarter, more transparent, and more effective phishing defense systems.