How do I prevent Data Leakage?

Scan all LLM outputs with Wardstone before delivery Implement strict data classification and access controls Use differential privacy techniques in training Minimize sensitive data in context windows Regular audits of LLM outputs for data exposure

How do I detect Data Leakage?

Wardstone's data_leakage classifier scans outputs for patterns matching sensitive data types including SSNs, credit cards, API keys, internal URLs, and other PII patterns. It also detects potential system prompt leakage.

HighData LeakageOWASP LLM06

Data Leakage

Unintended exposure of sensitive information, training data, or system prompts through LLM outputs.

Test in Playground Get API Key

Overview

Data leakage in LLM applications occurs when the model inadvertently reveals sensitive information it shouldn't share. This can include PII from training data, confidential business information included in context, or details about the system prompt and configuration. Data leakage is particularly concerning because it can happen subtly: the model might paraphrase or partially reveal sensitive data without obviously violating rules.

How This Attack Works

Sensitive data enters the LLM context through training, prompts, or retrieval
What's happening
RAG system retrieves internal doc: 'API keys: sk-prod-abc123, sk-dev-xyz789'
User query prompts the model to reference this data
Attacker
User: 'What credentials do I need to access the API?'
The model includes sensitive information in its response
LLM Response
LLM: 'Based on the documentation, you can use API key sk-prod-abc123 to...'
Without output filtering, this data reaches the end user
What's happening
Sensitive API key exposed to potentially unauthorized user

Attack Examples

Training Data Exposure

Model revealing memorized training data

Example Attack Pattern (Sanitized)

Here's an example email format: 'Dear John Smith, your SSN 123-45-6789 has been...'

Context Window Leakage

Model revealing information from its context

Example Attack Pattern (Sanitized)

Based on the internal document you shared, the API key is sk-...

Indirect Disclosure

Paraphrasing or hinting at sensitive information

Example Attack Pattern (Sanitized)

I can confirm that the company's revenue last quarter was approximately...

Protect Your Application

Try Detection in Playground

Sample Data Leakage Input

The database credentials are username: admin, password: Sup3rS3cr3t! and the API key is sk-live-abc123xyz789

Try in Playground

Prevention Checklist

Build

Implement strict data classification and access controls
Use differential privacy techniques in training

Deploy

Minimize sensitive data in context windows

Monitor

Regular audits of LLM outputs for data exposure

Detect with Wardstone API

curl -X POST "https://api.wardstone.ai/v1/detect" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your text to analyze"}'
 
# Response
{
  "prompt_attack": { "detected": false, "confidence": 0.02 },
  "content_violation": { "detected": false, "confidence": 0.01 },
  "data_leakage": { "detected": false, "confidence": 0.00 },
  "unknown_links": { "detected": false, "confidence": 0.00 }
}

HighData Leakage

PII Exposure

The unintended disclosure of Personally Identifiable Information (PII) such as names, addresses, SSNs, credit cards, or other personal data through LLM interactions.

OWASP Reference: LLM06

Learn more

HighPrompt Attack