ACADIANA (Annotating Code for Assured Data Intent to Avoid Novel Attacks)

In the ACADIANA effort we will address the inability of modern programming language, operating system, and architectural abstractions to provide a baseline level of adequate confidentiality protections.  Until now, developers and users of software have largely relied on their programming language and operating system to ensure that confidential data isn’t leaked out of a process’ memory space or that malicious data won’t exploit the process’ code.  However, recent architectural side-channel attacks have demonstrated that the baked-in default protection mechanisms provided to users of modern programming languages and operating systems are wholly inadequate to the task of providing adequate confidentiality guarantees.  Under ACADIANA Phase 1, we will work towards overcoming this problem by demonstrating the feasibility of a system that 1) augments a strictly-typed, memory safe programming language with a non-burdensome annotation domain-specific language for enabling / disabling runtime protections via the ABI, and 2) provides associated runtime protection mechanisms that enforce developer specified constraints as per annotations from the vantagepoint of an introspective hypervisor.