Inspiration
At Modus Create, we bake security into everything we do. Wouldn't it be cool if AI could help us identify malicious attacks as they happen, and LaunchDarkly could help us redirect those attackers to an isolated corner where they can't cause any more problems?
What it does
This app uses AI to identify malicious attack patterns and uses LaunchDarkly feature flags to redirect malicious traffic to a dummy landing page.
How we built it
We created a React front-end to handle requests. Each request is sent to our Python backend and validated through Bedrock to see if it includes any malicious attack vectors. If an attempted attack is detected, the LaunchDarkly API is used to update the flag status accordingly, thus harmlessly redirecting the attacker to a dummy page.
Challenges we ran into
Even with Bedrock in the mix, we still had to think about different payloads when using different AI models and handle the responses appropriately. It also took a bit to figure out how to update flag values via code.
Accomplishments that we're proud of
Being able to use AI is a very exciting process. Having the wide variety of models at your fingertips was very cool, and connecting to LaunchDarkly as a simple way to manage "permissions" was incredibly rewarding. Finally, pulling this all together in such a short amount of time felt like an accomplishment in itself!
What we learned
It was a first for us interacting w/ AI at a developer level. We were surprised to see the number of different models available and how the responses varied among them. We discovered that the more advanced models are clearly smarter but also much more expensive and resource heavy. Figuring out when cheaper models are more economical without detrimental trade-offs is key to success.
What's next for Modus Security
We'd love to explore turning this app into a reference architecture to inform future projects!
Log in or sign up for Devpost to join the conversation.