03. Tenant Isolation

Chapter 3 of 24 · 15 min

KEY INSIGHT

Tenant isolation goes beyond database separation—cache keys, file storage, external API calls, and logs must all respect tenant boundaries. Data isolation failures typically occur at system boundaries. The database query is properly filtered, but the cache returns another tenant's data, or the S3 bucket contains mixed tenant files. A defense-in-depth approach treats every data access point as a potential breach vector. Cache isolation prevents tenant data leakage in Redis or Memcached. Every cache key must include the tenant identifier. ```python import redis from functools import wraps redis_client = redis.Redis(host="localhost", port=6379, db=0) def tenant_cache(key_template: str, ttl: int = 300): """Decorator that prefixes cache keys with tenant ID.""" def decorator(func): @wraps(func) def wrapper(tenant_id: str, *args, **kwargs): cache_key = f"tenant:{tenant_id}:{key_template.format(*args, **kwargs)}" cached = redis_client.get(cache_key) if cached: return json.loads(cached) result = func(tenant_id, *args, **kwargs) redis_client.setex(cache_key, ttl, json.dumps(result)) return result return wrapper return decorator @tenant_cache("models:{model_id}", ttl=600) def get_model_config(tenant_id: str, model_id: str) -> dict: return db.query(ModelConfig).filter_by( tenant_id=tenant_id, id=model_id ).first() ``` File storage isolation matters when tenants upload files or when the system generates outputs. Object stores like S3 or MinIO require careful bucket and key design. ```python import boto3 from typing import BinaryIO class TenantFileStore: def __init__(self, bucket_name: str): self.s3 = boto3.client("s3") self.bucket = bucket_name def _validate_path(self, tenant_id: str, key: str): """Prevent path traversal attacks.""" if ".." in key or key.startswith("/"): raise ValueError("Invalid path") # Enforce tenant prefix expected_prefix = f"tenants/{tenant_id}/" if not key.startswith(expected_prefix): raise ValueError(f"Path must start with {expected_prefix}") def upload(self, tenant_id: str, key: str, file_obj: BinaryIO): self._validate_path(tenant_id, key) s3_key = f"tenants/{tenant_id}/{key}" self.s3.upload_fileobj(file_obj, self.bucket, s3_key) ``` Failure mode: omitting validation in file operations allows one tenant to overwrite another tenant's files or access sensitive data. The `_validate_path` function prevents directory traversal but a missing check anywhere in the upload/download chain creates vulnerability.

EXERCISE

Audit this code for isolation vulnerabilities. Write a test that demonstrates what happens when get_model_config is called with the wrong tenant_id. How would you add integration test coverage for tenant isolation?