phishpilot

Inspiration

Phishing emails and malicious URLs can trick people into sharing sensitive information. I wanted to build a tool that helps users analyse suspicious content and understand potential threats using AI

What it does

PhishPilot analyses suspicious URLs and emails using a local Decider service and Amazon Bedrock Nova Micro for AI-generated reports. It provides risk classifications, analysis evidence, and explanations to help users make informed decisions

How we built it

We built PhishPilot using Python and Streamlit, integrated a local Decider service for analysis, and connected Amazon Bedrock Nova Micro to generate detailed reports. The application combines local decision-making with cloud-based AI reporting

Challenges we ran into

Integrating the local Decider service with AWS Bedrock was a key challenge. Debugging service communication, handling AWS authentication, and managing uncertain analysis results helped us understand the practical complexities of building AI-powered cybersecurity applications

Accomplishments that we're proud of

We completed a working prototype that brings together URL and email analysis, local AI decision-making, and AI-generated security reports in one application

What we learned

We gained practical experience with Python, Streamlit, AWS Bedrock, service integration, and building a cybersecurity application that combines local processing with cloud AI

What's next for PhishPilot

We plan to improve analysis quality, strengthen evidence validation, enhance report generation, and explore additional indicators to make PhishPilot more useful for real-world phishing analysis

Built With

• agents
• amazon
• amazon-web-services
• bedrock
• boto3
• cybersecurity
• machine
• machine-learning
• python
• strands
• streamlit