Sentinel AI Guardian

The ultimate crytpocurrency wallet anti-blind signing solution with multi-chain support.

Comment

Inspiration

I was inspired to build this web security platform due to the ongoing crypto phishing scams and wallet drainers all over the web. Blind signing has been a serious problem over the years and many people "including friends of mine" have fallen victim to wallet drainers, crypto scams, and hacking attacks. In the world of evolving cyber threats it just made sense to build a tool that'll safeguard everyone operating and using crypto currency wallets/sites even though they have minimal technical experience.

What it does

It protects users from approving fraudulent transactions via their crypto currency wallets by alerting them of potential risks via the smart contract they're interacting with. It also alerts users of known phishing sites recorded to Alchemy or our local database. In total it protects users from blind signing transactions with support for multiple crypto chains. A chrome extension scans every wallet crypto transaction via an automated injection script for active analysis. The extension also connects to the active website a user is on and scans it for active threats

How we built it

Humans don't have the ability to decode raw call data often in Hexadecimals. The Amazon Bedrock model I integrated automatically decodes the transaction hex and smart contract functions before a user approves it. I built the web app in various phases:

  1. I scaffolded the frontend using the V0 portal and exported the project files to my local IDE.
  2. I designed the backend in the root folder of the web app with typescript and Node - The AWS powered backend engine supports scans via a remote injection trigger that works with the bundled chrome extension and the vercel deployment URL acting as the full service API.
  3. I deployed the frontend to Vercel and the Serverless Backend to AWS.
  4. I made use of the AWS DynamoDB for storing user dashboard and account data. All wallet data is stored on the NoSQL database.
  5. I integrated an AWS Bedrock Model "Claude Sonnet 4.5" for AI analysis.
  6. I performed a security audit and integrated advanced rate limiting for improved platform performance and protection.
  7. I designed the chrome extension using Plasmo and created a bundled version that is downloadable on the website with installation instructions.

Challenges we ran into

  1. It took a lot to get the injection timing right so that every crypto transaction is scanned correctly.
  2. The serverless backend gave a lot of issues during the early stages due to the multiple features I integrated.
  3. Mobile UI compatibility was also an issue in which I later solved.

Accomplishments that we're proud of

  1. The web app is fully up and running on Vercel and AWS.
  2. All preventive security features work as envisioned.
  3. The Chrome extension works seamlessly.
  4. The Amazon AWS Bedrock models and Dynamo database worked as I wanted them to in the web app.

What we learned

  1. I learned how to use Amazon's AWS Bedrock Models in an application.
  2. I learned how to create a Chrome extension.
  3. I learned a lot about onchain cybersecurity.

What's next for Sentinel AI Guardian

I'll be expanding the team and upgrading the services we provide to paid tiers. We'll be making use of a more advanced AWS backend with improved security detection techniques. The bundled extension would be deployed to the chrome web store and other web stores. The mobile app version would be worked on and deployed to Google's play store and Apple's app stores respectively.

Built With

Share this project:

Updates