Secured AI chatbot

Goal:

The purpose of this demonstration is to showcase the capability to identify Personally Identifiable Information (PII) and Personal Information (PI) within user input, mask it before transmitting the data to a Large Language Model (LLM) or other services that function without requiring personal details such as names, phone numbers, or bank account information, and subsequently reconstruct the data to present the final result to the user.

 

Features:

1. Detectable Entities: Credit Card, Email Address, IBAN Code, Person Name, Phone Number
2. Text Sources: User messages or uploaded files can be processed for entity detection.
3. Interactive Mode: Displays system operations, including how data is masked before being sent to the LLM, the LLM’s response, and the reconstructed message.

 

Challenges:

• Accurate Entity Detection: Ensuring high precision in identifying Personally Identifiable Information (PII) while minimizing false positives and negatives.
• Efficient Real-Time Masking: Maintaining low latency while masking sensitive data before sending it to external services, ensuring seamless user experience.
• Context-Aware Reconstruction: Rebuilding the masked information correctly without distorting the intended meaning or causing inconsistencies in the final response.