Scam Scanner

An AI model that will assist you in keeping you safe.

Comment

Inspiration

With the ever-evolving technology around us, one can always be sure that it may fall into the wrong hands. How can we be certain that we'll always stay one step ahead? What about the people who we personally know who did in fact fall victims to them? We created Scam Scanner with our relatives and friends in mind, who have been fooled in the past by fraudulent messages, due to unfamiliarity with different things such as the English language used in the messages, or the workings of how things work in the new country they've come to.

What it does

Scam Scanner is a scam detector for text-based scams, such as SMS, emails, and audio transcription of calls. It uses a predictive AI model that uses deep learning to detect the likelihood of the message being a fraudulent attempt.

How we built it

We understand that our target user base may be less accustomed to using technology, which could have been a reason for not being familiar with the signs of a scam. So, we tried to make our UI as simple and easy to use as possible, and this was accomplished with our prototype built using Figma. As we aim for user convenience above all, Scam Scanner takes the form of an application directly on the user’s phone that they can use anytime and anywhere. React Native is a Javascript-based mobile app framework that allows us to make our app cross-platform to be utilized on iOS and Android, while also allowing us in customizing our app to fit our desired design. Our AI is trained via a Recurrent Neural Network model handled through Python 3.

Challenges we ran into

While all members of the team are familiar with different programming languages, none of us had experience with React Native and this proved to be a challenge in finalizing our project and its form as an application. However, this was a great chance to try our hands at it and we hope to successfully use it in future projects!

It was also challenging to cut our pitch to under 2 minutes, but hopefully we were able to express our vision well enough in the final cut!

Accomplishments that we're proud of

Our datasets used to train our AI model helped detect scam messages with a high degree of accuracy of over 80% despite the short training time. With more time, we hope to achieve a high level of accuracy for our model.

What's next for Scam Scanner

Our next steps are to finish implementing our application in React Native, to test its functionality on physical mobile devices, while also applying the UI designed via HTML rendering. To ensure our app is comfortable to use and intuitive to a new user, we aim to conduct tests to assess different criteria of Scam Scanner, while also adding our additional planned features and their testing data such as phishing E-mail and call transcription datasets.

Built With

Share this project:

Updates