AI Privacy Firewall for Java

Intercept LLM prompts, detect PII using a hybrid Regex + Naive Bayes pipeline, and replace sensitive data with structured tokens — before it leaves your network.

SSN Credit Card Email Phone IPv4 / IPv6 API Keys Passwords Names (ML) Orgs (ML) IBAN

Input Text

Mode: ML threshold:

Redacted Output

🔐

Redacted text will appear here

Reverse Map (de-tokenisation)

How It Works

Input

Raw Text

Any string: LLM prompt, email body, support ticket, log line…

Layer 1

HeuristicDetector

10 pre-compiled regexes. Luhn for cards, NANP for phones, entropy filter for API keys.

Layer 2

MLDetector

Multinomial Naive Bayes. Classifies title-cased tokens using bag-of-words context features.

Merge

CompositeDetector

De-duplicates overlapping matches. Promotes same-span matches to HYBRID source.

Output

PIITokenizer

TOKEN / MASK / BLANK modes. Reverse map for de-tokenisation after LLM response.