Session management from a code review perspective should focus on the creation, renewal and destruction of a users session throughout the application. The code review process should ensure the following:
- Authenticated users have a robust and cryptographically secure association with their session.
- The session identifier (Session ID) shoud not be predictable and generation of such should be left to the underlaying framework. - The development effort to produce a session with sufficient entrophy is subject to errors and best left to tried and trusted methods.
- Applications enforce authorization checks
- Applications should check if the session is valid prior to servicing any user requests. The users session object may also hold authorisation data.
- Session ID should be applied to a new user upon successful authentication.
- Reviewing the code to identify where sessions are created and invalidated is important.
- Sessions may need to be terminated upon authorization failures. If a logical condition exists which is not pssible unless the state transition is circumvented or an obvious attempt to escalate privlidges, a session should be terminated.
- Applications avoid or prevent common web attacks, such as replay, request forging and man-in-the-middle.
Session identifiers should be passed to the user in a secure manner such as not using HTTP GET with the session ID being placed in the query string. - Such data (query string) is logged in web server logs.
- Session Timeout- Sessions should have a defined inactivity timeout and also in some cases a session hardlimit. The code review should examine such session settings. They may be defined in configuration files or in the code itself. Hard limits shall kill a session regardless of session activity.