Inspiration
We felt that as the world became More and more digital, the importance of cybersecurity becomes more prevalent , and we felt that it could interesting to see how can we use or if the Use of AI in cybersecurity could vastly reduce the risks of fraud.
What it does
It basically reads in a Transaction statement of a user and compares that transaction statement to what the users previous transaction statements were in order to accurate predict if its a valid or invalid transaction
How we built it
We used the vast services of amazon AWS to build an AI that has the users previous transaction statements, we then utilise an agent to determine if the transaction statement is valid by giving certain weights to each fields, eg:- amount, Merchant Name,Location
Challenges we ran into
While we were able to build the AI we ran into the problem with the interface parts of the problem as we were unable to integrate it with an output surface. A few other problems we ran into was the Use of the YAML language in AWS as it was completely new thing for us, we also ran into a few debugging while training the AI to get a desired Output.
Accomplishments that we're proud of
While we weren't able to produce a full stack solution for this, we are pretty ecstatic that we as a team who have little to no experience building an actual AI bot were able to build the system with our own skills and despite all the challenges we faced we pushed through and were able to get an actual working AI system that detects fraud
What we learned
We learned a lot about the vast services AWS provides and how its a useful tool when it comes to building projects and how its such a versatile platform, We also learned about how cybersecurity and AI could be used together.
What's next for CreditCardFraudDetector
The future for this project I think could be we actually build an app that Would take in the SMS messages from your phones Put it through the AI and determine if the transaction is fraud or not. If so we could then use it to block any payments to those certain services, We could do this on a large scale and actually build a database of vendors who would have a Fraudprobability score where the lowest is them being least fraudulent and the highest being the Most.
Log in or sign up for Devpost to join the conversation.