Experiments and Proofs in Web-service Security

Many web services have a subsystem for allowing users to register, authenticate, reset their password, and change personal details. It is important that such subsystems cannot be abused by attackers to gain access to the accounts of other users. We study a system which was initially prone to such attacks. Specific attacks are demonstrated and the system is then modified to prevent such attacks in future. The design achieved in this way is then analysed to show that it can't be broken in future unless users allow their email to be intercepted. This is achieved by formulating the requirement as a statement of the user's expectations of the system and then analysing the source code of the system to prove that it meets these requirements. The process of attack, correction, and formulation of security rules, and proof that rules hold, is proposed as a methodical security design philosophy.