When implementing security measures, always err on the side of caution by making conservative assumptions about potential risks. This principle applies across multiple security contexts: **Taint Analysis**: If there's any execution path where data could become tainted, treat it as tainted. Use may-analysis approaches that assume the worst-case scenario rather than trying to prove safety. **Capability Restriction**: Limit access rights to the absolute minimum necessary. Don't pass unnecessary capabilities to callbacks or sub-components, even if they're currently unused. **Resource Protection**: Implement defensive measures like timeouts to prevent resource exhaustion, even in contexts where attacks seem unlikely. Example from taint analysis: ```python def process_data(user_input): try: safe_data = validate(user_input) result = process(safe_data) except ValidationError: result = user_input # This path taints result # Treat result as tainted due to exception path return sanitize(result) ``` This conservative approach helps prevent security vulnerabilities by assuming potential attack vectors exist rather than trying to prove they don't.