AMAZESECURE

AmazeSecure AI protects shipments with secure QR labels, LLMs, and anomaly detection, revealing data only when safe to ensure privacy, security, and trust.

Comment

AmazeSecure AI

AmazeSecure AI is an autonomous AI agent built on AWS to protect sensitive shipping information using QR-based smart labels, reasoning LLMs, and anomaly detection.

Instead of printing addresses, names, or phone numbers on packages, each package has a secure QR code linked to a package_id.
The AI agent uses Large Language Models (LLMs) to reason in real-time whether to reveal the data, detect suspicious activity, and generate alerts to ensure privacy, security, and trust in deliveries.

Inspiration

Every package we receive contains sensitive information: full name, address, phone number, and sometimes additional order details. Once discarded, these physical labels become a tangible vector for identity theft and social engineering fraud. Criminals can use this data to send fake messages, create fake identities, or even attempt physical scams by impersonating delivery personnel.

While most identity theft originates from digital breaches, practical evidence and consumer protection warnings show that discarded physical information is far from harmless. In Mexico and worldwide, discarded shipping labels represent a potential risk that rarely has technological solutions.

AmazseSecure was developed by me independently to protect personal information from delivery to final package disposal. The project replaces or complements printed sensitive data with a temporary QR code, accessible only for a limited time and to authorized personnel, reducing exposure and preventing both physical and digital fraud linked to traditional labels. This approach not only protects consumers but also strengthens trust in e-commerce and last-mile logistics … and also shows that the customer is a priority at every step.

I am passionate about Artificial Intelligence, and as a final-year Computer Science student, this project also reflects my motivation to prepare for an internship within the next 8 months.

What I Learned

  • How to integrate multiple AWS services such as Bedrock, SageMaker, KMS, and DynamoDB into a single, secure workflow.
  • The importance of combining reasoning LLMs with anomaly detection for real-world security scenarios.
  • How to turn theoretical AI concepts into a working prototype that solves a real privacy problem.

How I Built It

  • Smart QR labels contain only a package_id.
  • A Lambda function receives requests from API Gateway, queries Bedrock AgentCore for reasoning with LLMs, and SageMaker for anomaly detection.
  • KMS decrypts sensitive data only when the request is validated.
  • SNS sends alerts for unauthorized access or suspicious activity.
  • CloudWatch logs all attempts for auditing and transparency.

Workflow

  1. Delivery personnel scans QR → receives package_id.
  2. App sends package_id → API Gateway → Lambda function.
  3. Lambda calls Bedrock + SageMaker:
    • Authorized user → KMS decrypts and returns the address.
    • Unauthorized user → access denied with explanation.
    • Suspicious activity → SNS alert triggered.
  4. CloudWatch logs all attempts.

Challenges

  • Building the full architecture and implementation independently.
  • Ensuring the idea was technically sound and practically usable.
  • Learning new AWS integrations while applying AI concepts in a real project.

The hardest part was balancing security, reasoning with LLMs, and usability, so delivery personnel could still operate smoothly.

Tech Stack

  • Language: Python 3.11
  • AWS Services:
    • Amazon Bedrock (LLM + AgentCore)
    • Amazon SageMaker (Anomaly Detection)
    • AWS Lambda (backend logic)
    • Amazon API Gateway (QR scan endpoint)
    • AWS KMS (encryption/decryption)
    • Amazon DynamoDB (package storage)
    • Amazon S3 (QR codes + encrypted data)
    • Amazon SNS (alerts)
    • Amazon CloudWatch (logging & auditing)
    • (Optional) Amazon Q and Nova for advanced integrations
  • Libraries: boto3, json, os

Deployment Steps

  1. Create S3 buckets for QR codes and encrypted data.
  2. Create DynamoDB table AmazeSecurePackages.
  3. Set up a KMS key for encryption/decryption.
  4. Create SNS topic for alerts.
  5. Deploy Lambda function with environment variables.
  6. Connect API Gateway to Lambda for QR scanning.
  7. (Optional) Train & deploy anomaly detection model on SageMaker.
  8. (Optional) Integrate with Bedrock AgentCore primitives, Amazon Q, and Nova.

Demo

The demo (to be added after Bedrock auto-activation on October 8, 2025) will show:

  • QR scanned → valid token → address revealed.
  • QR scanned → invalid/expired token → SNS alert.
  • Multiple failed scans → anomaly detection → alert triggered and access blocked.

Math Support Example

Security decisions can be expressed as a probability model:

[ P(\text{Access Granted}) = P(\text{Valid Token}) \times P(\text{Authorized User}) \times P(\text{No Anomaly}) ]

Final Reflection

This project gave me confidence in building end-to-end AI-powered systems on the cloud.
I believe AmazeSecure AI can help create trustworthy, secure, and private deliveries, while giving me real-world skills in AI, cloud computing, and data security.

⚠️ Note: The system is intended to be powered by Bedrock LLMs, combining reasoning with anomaly detection for secure decision-making. Integration with advanced AI services like AgentCore, Amazon Q, and Nova is planned to enhance automation and real-time threat analysis.

Built With

python · lambda · boto3 · amazon-bedrock · amazon-sagemaker · api-gateway · awskms · amazon-dynamodb · amazon-s3 · amazon-sns · amazon-cloudwatch · amazon-q

Built With

Share this project:

Updates