How to use Weave with PII data:
In this tutorial, we'll demonstrate how to utilize Weave while ensuring your Personally Identifiable Information (PII) data remains private. Weave supports removing PII from LLM calls and preventing PII from being displayed in the Weave UI.
To detect and protect our PII data, we'll identify and redact PII data and optionally anonymize it with the following methods:
- Regular expressions to identify PII data and redact it.
- Microsoft's Presidio, a python-based data protection SDK. This tool provides redaction and replacement functionalities.
- Faker, a Python library to generate fake data, combined with Presidio to anonymize PII data.
Additionally, we'll make use of Weave Ops input/output logging customization to seamlessly integrate PII redaction and anonymization into the workflow. See here for more information.
For this use-case, we will leverage Anthropic's Claude Sonnet to perform sentiment analysis while tracing the LLM calls using Weave's Traces. Sonnet will receive a block of text and output one of the following sentiment classifications: positive, negative, or neutral.
Overview of Weave Ops Input/Output Logging Customization
Weave Ops support defining input and output postprocessing functions. These functions allow you to modify the data that is passed to your LLM call or logged to Weave, respectively.
from dataclasses import dataclass
from typing import Any
import weave
# Inputs Wrapper Class
@dataclass
class CustomObject:
x: int
secret_password: str
# First we define functions for input and output postprocessing:
def postprocess_inputs(inputs: dict[str, Any]) -> dict[str, Any]:
return {k:v for k,v in inputs.items() if k != "hide_me"}
def postprocess_output(output: CustomObject) -> CustomObject:
return CustomObject(x=output.x, secret_password="REDACTED")
# Then, when we use the `@weave.op` decorator, we pass these processing functions as arguments to the decorator:
@weave.op(
postprocess_inputs=postprocess_inputs,
postprocess_output=postprocess_output,
)
def some_llm_call(a: int, hide_me: str) -> CustomObject:
return CustomObject(x=a, secret_password=hide_me)
Setup
Let's install the required packages and set up our API keys. Your Weights & Biases API key can be found here, and your Anthropic API keys are here.
%%capture
# @title required python packages:
!pip install presidio_analyzer
!pip install presidio_anonymizer
!python -m spacy download en_core_web_lg # Presidio uses spacy NLP engine
!pip install Faker # we'll use Faker to replace PII data with fake data
!pip install weave # To leverage Traces
!pip install set-env-colab-kaggle-dotenv -q # for env var
!pip install anthropic # to use sonnet
!pip install cryptography # to encrypt our data
%%capture
# @title Make sure to set up set up your API keys correctly
# See: https://pypi.org/project/set-env-colab-kaggle-dotenv/ for usage instructions.
from set_env import set_env
_ = set_env("ANTHROPIC_API_KEY")
_ = set_env("WANDB_API_KEY")
import weave
# Start a new Weave project
WEAVE_PROJECT = "pii_cookbook"
weave.init(WEAVE_PROJECT)
Let's load our initial PII data. For demonstration purposes, we'll use a dataset containing 10 text blocks. A larger dataset with 1000 entries is available.
import requests
url = "https://raw.githubusercontent.com/wandb/weave/master/docs/notebooks/10_pii_data.json"
response = requests.get(url)
pii_data = response.json()
print('PII data first sample: "' + pii_data[0]["text"] + '"')
Redaction Methods Implementation
Method 1: Regular Expression Filtering
Our initial method is to use regular expressions (regex) to identify PII data and redact it. It allows us to define patterns that can match various formats of sensitive information like phone numbers, email addresses, and social security numbers. By using regex, we can scan through large volumes of text and replace or redact information without the need for more complex NLP techniques.
import re
# Define a function to clean PII data using regex
def redact_with_regex(text):
# Phone number pattern
# \b : Word boundary
# \d{3} : Exactly 3 digits
# [-.]? : Optional hyphen or dot
# \d{3} : Another 3 digits
# [-.]? : Optional hyphen or dot
# \d{4} : Exactly 4 digits
# \b : Word boundary
text = re.sub(r"\b\d{3}[-.]?\d{3}[-.]?\d{4}\b", "<PHONE>", text)
# Email pattern
# \b : Word boundary
# [A-Za-z0-9._%+-]+ : One or more characters that can be in an email username
# @ : Literal @ symbol
# [A-Za-z0-9.-]+ : One or more characters that can be in a domain name
# \. : Literal dot
# [A-Z|a-z]{2,} : Two or more uppercase or lowercase letters (TLD)
# \b : Word boundary
text = re.sub(
r"\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b", "<EMAIL>", text
)
# SSN pattern
# \b : Word boundary
# \d{3} : Exactly 3 digits
# - : Literal hyphen
# \d{2} : Exactly 2 digits
# - : Literal hyphen
# \d{4} : Exactly 4 digits
# \b : Word boundary
text = re.sub(r"\b\d{3}-\d{2}-\d{4}\b", "<SSN>", text)
# Simple name pattern (this is not comprehensive)
# \b : Word boundary
# [A-Z] : One uppercase letter
# [a-z]+ : One or more lowercase letters
# \s : One whitespace character
# [A-Z] : One uppercase letter
# [a-z]+ : One or more lowercase letters
# \b : Word boundary
text = re.sub(r"\b[A-Z][a-z]+ [A-Z][a-z]+\b", "<NAME>", text)
return text
Let's test the function with a sample text:
# Test the function
test_text = "My name is John Doe, my email is john.doe@example.com, my phone is 123-456-7890, and my SSN is 123-45-6789."
cleaned_text = redact_with_regex(test_text)
print(f"Raw text:\n\t{test_text}")
print(f"Redacted text:\n\t{cleaned_text}")
Method 2: Microsoft Presidio Redaction
Our next method involves complete removal of PII data using Presidio. This approach redacts PII and replaces it with a placeholder representing the PII type.
For example:
"My name is Alex" becomes "My name is <PERSON>".
Presidio comes with a built-in list of recognizable entities. We can select the ones that are important for our use case. In the below example, we redact names, phone numbers, locations, email addresses, and US Social Security Numbers.
We'll then encapsulate the Presidio process into a function.
from presidio_analyzer import AnalyzerEngine
from presidio_anonymizer import AnonymizerEngine
# Set up the Analyzer, which loads an NLP module (spaCy model by default) and other PII recognizers.
analyzer = AnalyzerEngine()
# Set up the Anonymizer, which will use the analyzer results to anonymize the text.
anonymizer = AnonymizerEngine()
# Encapsulate the Presidio redaction process into a function
def redact_with_presidio(text):
# Analyze the text to identify PII data
results = analyzer.analyze(
text=text,
entities=["PHONE_NUMBER", "PERSON", "LOCATION", "EMAIL_ADDRESS", "US_SSN"],
language="en",
)
# Anonymize the identified PII data
anonymized_text = anonymizer.anonymize(text=text, analyzer_results=results)
return anonymized_text.text
Let's test the function with a sample text:
text = "My phone number is 212-555-5555 and my name is alex"
# Test the function
anonymized_text = redact_with_presidio(text)
print(f"Raw text:\n\t{text}")
print(f"Redacted text:\n\t{anonymized_text}")
Method 3: Anonymization with Replacement using Fakr and Presidio
Instead of redacting text, we can anonymize it by swapping PII (like names and phone numbers) with fake data generated using the Faker Python library. For example:
"My name is Raphael and I like to fish. My phone number is 212-555-5555"
might become
"My name is Katherine Dixon and I like to fish. My phone number is 667.431.7379"
To effectively utilize Presidio, we must supply references to our custom operators. These operators will direct Presidio to the functions responsible for swapping PII with fake data.
from faker import Faker
from presidio_anonymizer import AnonymizerEngine
from presidio_anonymizer.entities import OperatorConfig
fake = Faker()
# Create faker functions (note that it has to receive a value)
def fake_name(x):
return fake.name()
def fake_number(x):
return fake.phone_number()
# Create custom operator for the PERSON and PHONE_NUMBER" entities
operators = {
"PERSON": OperatorConfig("custom", {"lambda": fake_name}),
"PHONE_NUMBER": OperatorConfig("custom", {"lambda": fake_number}),
}
text_to_anonymize = (
"My name is Raphael and I like to fish. My phone number is 212-555-5555"
)
# Analyzer output
analyzer_results = analyzer.analyze(
text=text_to_anonymize, entities=["PHONE_NUMBER", "PERSON"], language="en"
)
anonymizer = AnonymizerEngine()
# do not forget to pass the operators from above to the anonymizer
anonymized_results = anonymizer.anonymize(
text=text_to_anonymize, analyzer_results=analyzer_results, operators=operators
)
print(f"Raw text:\n\t{text_to_anonymize}")
print(f"Anonymized text:\n\t{anonymized_results.text}")
Let's consolidate our code into a single class and expand the list of entities to include the additional ones we identified earlier.
from faker import Faker
from presidio_anonymizer import AnonymizerEngine
from presidio_anonymizer.entities import OperatorConfig
# A custom class for generating fake data that extends Faker
class my_faker(Faker):
# Create faker functions (note that it has to receive a value)
def fake_address(x):
return fake.address()
def fake_ssn(x):
return fake.ssn()
def fake_name(x):
return fake.name()
def fake_number(x):
return fake.phone_number()
def fake_email(x):
return fake.email()
# Create custom operators for the entities
operators = {
"PERSON": OperatorConfig("custom", {"lambda": fake_name}),
"PHONE_NUMBER": OperatorConfig("custom", {"lambda": fake_number}),
"EMAIL_ADDRESS": OperatorConfig("custom", {"lambda": fake_email}),
"LOCATION": OperatorConfig("custom", {"lambda": fake_address}),
"US_SSN": OperatorConfig("custom", {"lambda": fake_ssn}),
}
def redact_and_anonymize_with_faker(self, text):
anonymizer = AnonymizerEngine()
analyzer_results = analyzer.analyze(
text=text,
entities=["PHONE_NUMBER", "PERSON", "LOCATION", "EMAIL_ADDRESS", "US_SSN"],
language="en",
)
anonymized_results = anonymizer.anonymize(
text=text, analyzer_results=analyzer_results, operators=self.operators
)
return anonymized_results.text
Let's test the function with a sample text:
faker = my_faker()
text_to_anonymize = (
"My name is Raphael and I like to fish. My phone number is 212-555-5555"
)
anonymized_text = faker.redact_and_anonymize_with_faker(text_to_anonymize)
print(f"Raw text:\n\t{text_to_anonymize}")
print(f"Anonymized text:\n\t{anonymized_text}")
Applying the Methods to Weave Calls
In these examples we will integrate our PII redaction and anonymization methods into Weave Models, and preview the results in Weave Traces.
We'll create a Weave Model which is a combination of data (which can include configuration, trained model weights, or other information) and code that defines how the model operates.
In this model, we will include our predict function where the Anthropic API will be called. Additionally, we will include our postprocessing functions to ensure that our PII data is redacted or anonymized before it is sent to the LLM.
Once you run this code you will receive a links to the Weave project page as well as the specific trace (LLM calls)you ran.
Regex Method
In the simplest case, we can use regex to identify and redact PII data in the original text.
import json
from typing import Any
import anthropic
import weave
# Define an input postprocessing function that applies our regex redaction for the model prediction Weave Op
def postprocess_inputs_regex(inputs: dict[str, Any]) -> dict:
inputs["text_block"] = redact_with_regex(inputs["text_block"])
return inputs
# Weave model / predict function
class sentiment_analysis_regex_pii_model(weave.Model):
model_name: str
system_prompt: str
temperature: int
@weave.op(
postprocess_inputs=postprocess_inputs_regex,
)
async def predict(self, text_block: str) -> dict:
client = anthropic.AsyncAnthropic()
response = await client.messages.create(
max_tokens=1024,
model=self.model_name,
system=self.system_prompt,
messages=[
{"role": "user", "content": [{"type": "text", "text": text_block}]}
],
)
result = response.content[0].text
if result is None:
raise ValueError("No response from model")
parsed = json.loads(result)
return parsed
# create our LLM model with a system prompt
model = sentiment_analysis_regex_pii_model(
name="claude-3-sonnet",
model_name="claude-3-5-sonnet-20240620",
system_prompt='You are a Sentiment Analysis classifier. You will be classifying text based on their sentiment. Your input will be a block of text. You will answer with one the following rating option["positive", "negative", "neutral"]. Your answer should be one word in json format: {classification}. Ensure that it is valid JSON.',
temperature=0,
)
print("Model: ", model)
# for every block of text, anonymized first and then predict
for entry in pii_data:
await model.predict(entry["text"])
Presidio Redaction Method
Here we will use Presidio to identify and redact PII data in the original text.
import json
from typing import Any
import anthropic
import weave
# Define an input postprocessing function that applies our Presidio redaction for the model prediction Weave Op
def postprocess_inputs_presidio(inputs: dict[str, Any]) -> dict:
inputs["text_block"] = redact_with_presidio(inputs["text_block"])
return inputs
# Weave model / predict function
class sentiment_analysis_presidio_pii_model(weave.Model):
model_name: str
system_prompt: str
temperature: int
@weave.op(
postprocess_inputs=postprocess_inputs_presidio,
)
async def predict(self, text_block: str) -> dict:
client = anthropic.AsyncAnthropic()
response = await client.messages.create(
max_tokens=1024,
model=self.model_name,
system=self.system_prompt,
messages=[
{"role": "user", "content": [{"type": "text", "text": text_block}]}
],
)
result = response.content[0].text
if result is None:
raise ValueError("No response from model")
parsed = json.loads(result)
return parsed
# create our LLM model with a system prompt
model = sentiment_analysis_presidio_pii_model(
name="claude-3-sonnet",
model_name="claude-3-5-sonnet-20240620",
system_prompt='You are a Sentiment Analysis classifier. You will be classifying text based on their sentiment. Your input will be a block of text. You will answer with one the following rating option["positive", "negative", "neutral"]. Your answer should be one word in json format: {classification}. Ensure that it is valid JSON.',
temperature=0,
)
print("Model: ", model)
# for every block of text, anonymized first and then predict
for entry in pii_data:
await model.predict(entry["text"])
Faker + Presidio Replacement Method
Here we will have Faker generate anonymized replacement PII data and use Presidio to identify and replace the PII data in the original text.
import json
from typing import Any
import anthropic
import weave
# Define an input postprocessing function that applies our Faker anonymization and Presidio redaction for the model prediction Weave Op
faker = my_faker()
def postprocess_inputs_faker(inputs: dict[str, Any]) -> dict:
inputs["text_block"] = faker.redact_and_anonymize_with_faker(inputs["text_block"])
return inputs
# Weave model / predict function
class sentiment_analysis_faker_pii_model(weave.Model):
model_name: str
system_prompt: str
temperature: int
@weave.op(
postprocess_inputs=postprocess_inputs_faker,
)
async def predict(self, text_block: str) -> dict:
client = anthropic.AsyncAnthropic()
response = await client.messages.create(
max_tokens=1024,
model=self.model_name,
system=self.system_prompt,
messages=[
{"role": "user", "content": [{"type": "text", "text": text_block}]}
],
)
result = response.content[0].text
if result is None:
raise ValueError("No response from model")
parsed = json.loads(result)
return parsed
# create our LLM model with a system prompt
model = sentiment_analysis_faker_pii_model(
name="claude-3-sonnet",
model_name="claude-3-5-sonnet-20240620",
system_prompt='You are a Sentiment Analysis classifier. You will be classifying text based on their sentiment. Your input will be a block of text. You will answer with one the following rating option["positive", "negative", "neutral"]. Your answer should be one word in json format: {classification}. Ensure that it is valid JSON.',
temperature=0,
)
print("Model: ", model)
# for every block of text, anonymized first and then predict
for entry in pii_data:
await model.predict(entry["text"])
Checklist for Safely Using Weave with PII Data
During Testing
- Log anonymized data to check PII detection
- Track PII handling processes with Weave Traces
- Measure anonymization performance without exposing real PII
In Production
- Never log raw PII
- Encrypt sensitive fields before logging
Encryption Tips
- Use reversible encryption for data you need to decrypt later
- Apply one-way hashing for unique IDs you don't need to reverse
- Consider specialized encryption for data you need to analyze while encrypted
Details
(Optional) Encrypting our data
In addition to anonymizing PII, we can add an extra layer of security by encrypting our data using the cryptography library's Fernet symmetric encryption. This approach ensures that even if the anonymized data is intercepted, it remains unreadable without the encryption key.
import os
from cryptography.fernet import Fernet
from pydantic import BaseModel, ValidationInfo, model_validator
def get_fernet_key():
# Check if the key exists in environment variables
key = os.environ.get('FERNET_KEY')
if key is None:
# If the key doesn't exist, generate a new one
key = Fernet.generate_key()
# Save the key to an environment variable
os.environ['FERNET_KEY'] = key.decode()
else:
# If the key exists, ensure it's in bytes
key = key.encode()
return key
cipher_suite = Fernet(get_fernet_key())
class EncryptedSentimentAnalysisInput(BaseModel):
encrypted_text: str = None
@model_validator(mode="before")
def encrypt_fields(cls, values):
if "text" in values and values["text"] is not None:
values["encrypted_text"] = cipher_suite.encrypt(values["text"].encode()).decode()
del values["text"]
return values
@property
def text(self):
if self.encrypted_text:
return cipher_suite.decrypt(self.encrypted_text.encode()).decode()
return None
@text.setter
def text(self, value):
self.encrypted_text = cipher_suite.encrypt(str(value).encode()).decode()
@classmethod
def encrypt(cls, text: str):
return cls(text=text)
def decrypt(self):
return self.text
# Modified sentiment_analysis_model to use the new EncryptedSentimentAnalysisInput
class sentiment_analysis_model(weave.Model):
model_name: str
system_prompt: str
temperature: int
@weave.op()
async def predict(self, encrypted_input: EncryptedSentimentAnalysisInput) -> dict:
client = AsyncAnthropic()
decrypted_text = encrypted_input.decrypt() # We use the custom class to decrypt the text
response = await client.messages.create(
max_tokens=1024,
model=self.model_name,
system=self.system_prompt,
messages=[
{ "role": "user",
"content":[
{
"type": "text",
"text": decrypted_text
}
]
}
]
)
result = response.content[0].text
if result is None:
raise ValueError("No response from model")
parsed = json.loads(result)
return parsed
model = sentiment_analysis_model(
name="claude-3-sonnet",
model_name="claude-3-5-sonnet-20240620",
system_prompt="You are a Sentiment Analysis classifier. You will be classifying text based on their sentiment. Your input will be a block of text. You will answer with one the following rating option[\"positive\", \"negative\", \"neutral\"]. Your answer should one word in json format dict where the key is classification.",
temperature=0
)
for entry in pii_data:
encrypted_input = EncryptedSentimentAnalysisInput.encrypt(entry["text"])
await model.predict(encrypted_input)