Modern multicore and manycore architectures have a number of new security threats. For example, shared microarchitecture components such as caches, core inter-connection networks, and memory controllers can be exploited for side-channel attacks or denial of service attacks. The evolution of workloads to exploit explicit parallelism will also likely lead to additional new threats. New forms of active viruses and trojans that reside on some cores and attempt to attack other applications are likely to arise. It is critical to anticipate such new forms of threats and design manycore systems in a manner that facilitates defeating them. If security is not treated as a first order design principle, these systems will be highly vulnerable to attacks, leading to enormous losses in money and productivity and a substantial effort to retrofit security in after the fact. The overall theme of this project is to identify and analyze security threats that can arise in a multicore and manycore environment and develop algorithms and techniques to address these threats in a complexity-effective manner and without sacrificing performance. Specific interest is on the techniques and solution patterns that can be reused to help with different threats. These include the use of additional cores and thread contexts to provide security without significant performance losses, and the development of techniques for virtual and physical isolation of shared resources to defend against sidechannel attacks and denial-of-service attacks. In addition, the project explores approaches for trading-off performance and security and also considers security vulnerabilities of new memory technologies.
SHF: Small: Architectural Support for Security in the Many-core Age: Threats and Opportunities