Frauducation

Inspiration

Our project - Frauducation - was inspired by the fact that many foreign students in the UK fall victim to fraud at a far higher rate than normal. The university provides a lecture at the beginning of the year highlighting the depth of the problem; however, the attendance is limited and the problem is persistent, thus we created Frauducation.

What it does

Frauducation is an online learning platform targeted at foreign university students. It provides a zone to access information on how to spot fraud, quizzes to help reinforce this information, and cutting-edge custom chatbots to simulate realistic fraud calls and messages. These chatbots interact with the users and provide different experiences depending on what it is more probable to be exposed to. We also have a reflection mechanism generated with AI that studies each user's results (Quiz answer, chat conversation...) and provides custom feedback based on that. We also developed a text-to-speech integration of the app, allowing for accessibility and the option of having more realistic call-like conversations.

How we built it

The front end was initially designed in Figma by our lead designer Lorenzo, before it was adapted into a traditional web interface built with node.js by our front-end engineer Thomas. On the backend, Edward built a flask-based API to communicate with the front end and integrate with models. Our AI engineer Alex used new tools from OpenAI’s GPT API to create custom fraud-bots and Ben built a bank of fraud information and a series of quiz questions based on common cases of fraud in our target audience.

Challenges we ran into

On the AI side of things, GPT is designed to be almost impossible to simulate fraud, so it was a huge task of trial and error to build an effective simulator for the bot.

Due to the fact that he was our only front-end engineer, Thomas had to build an entire website in 24 hours based on ambitious designs which was an extreme feat.

Both Benitas and Edward experienced numerous difficulties balancing the front and back-end and keeping it efficient, especially with the painful nature of JSONs.

Accomplishments that we're proud of

• Getting GPT to successfully break its boundaries and emulate fraudulent scenarios.
• Having a decent design while being a team of Developers
• Accumulating concise and relevant information that we believe will help students' security significantly
• Built a genuinely good-looking GUI with limited resources and an even more limited time frame.

What we learned

We've all gained a lot of experience working with web development both front and back end, as well as collaborating as a team. We've learnt how important planning everything is and through Figma & Canva designing and planning software we improved our workflow massively. We learned the importance of collaboration and constant communication, putting the right channels in place for that.

What's next for Frauducation

We genuinely want to propose our working website to the University of Exeter international student team (We identified some main points of contact already). We plan to put in place a close collaboration with them to hopefully have a product ready for next year to be distributed to international students. If it works, we plan on expanding to other universities and later even to other target users and maybe even businesses, thinking about fraud education for employees.

Built With

• css
• flask
• gpt3-turbo
• javascript
• python
• react