ScamSim

Agentic Voice and Email Cyber Resilience Training Platform

Group Name: Syntax Error

Inspiration

The growing number of phishing and voice scams shows that traditional cybersecurity awareness training is no longer enough. We were motivated to design a solution that allows employees to safely practise responding to realistic scam situations before encountering them in real life.

What it does

ScamSim is a proposed cyber resilience training platform that simulates realistic email and voice scam scenarios. It helps employees practise identifying threats and responding correctly under pressure in a controlled environment, turning awareness into hands-on preparedness.

How we plan to build it

We plan to develop a Python-based desktop application using:

  • Pygame for the visual interface (mandatory)
  • LangChain for agent orchestration and AI workflows
  • Azure OpenAI via LangChain for text generation and evaluation
  • Pydantic to structure and validate agent state and responses
  • Beautiful Soup to scrape real-world scam data for scenario grounding
  • ElevenLabs for realistic voice scam simulation

This stack ensures ScamSim is interactive, agent-driven, and grounded in real-world scams.

Challenges we expect to face

We anticipate challenges in designing convincing voice scam interactions, managing multi-step agent workflows, and balancing realism with safe and ethical training practices.

Accomplishments we aim to achieve

Our goal is to produce a working prototype that demonstrates agent-driven scam simulations across email and voice channels, with adaptive difficulty and actionable feedback for learning.

What we expect to learn

Through this project, we aim to gain hands-on experience in agentic AI design, multi-modal interaction systems, and creating practical cybersecurity training tools for real-world organisational use.

What’s next for Syntax Error

Next, we plan to refine scenario realism, expand adaptive difficulty levels, and explore deployment options for organisational cybersecurity training programmes.

Built With

Share this project:

Updates