Microsoft software security expert Michael Howard defines the following heuristics for determining code review in "A Process for Performing Security Code Reviews": Old code:
Newer code provides better understanding of software security and has lesser number of vulnerabilities. Older code must be checked deeply. Code that runs by default: It must have high quality, and must be checked deeply than code that does not execute by default. Code that runs by default increases the application's attack surface.
Code that runs in elevated context: It must have higher quality. Code that runs in elevated privileges must be checked deeply and increases the application's attack surface. Anonymously accessible code: It must be checked deeply than code that only authorized users and administrators can access, and it increases the application's attack surface. Code listening on a globally accessible network interface: It must be checked deeply for security vulnerabilities and increases the application's attack surface. Code written in C/C++/assembly language: It is prone to security vulnerabilities, for example, buffer overruns. Code with a history of security vulnerabilities: It includes additional vulnerabilities except concerted efforts that are required for removing them. Code that handles sensitive data: It must be checked deeply to ensure that data is protected from unintentional disclosure. Complex code: It includes undiscovered errors because it is more difficult to analyze complex code manually and programmatically. Code that changes frequently: It has more security vulnerabilities than code that does not change frequently.