"Is a Security Feature on the Way That Makes Computing Faster?"

The multiple programs running simultaneously on a device depend on data stored in the device's memory hardware. However, sensitive information may not be shared among all the programs, thus leaving the device exposed to a memory timing side-channel attack. When attempts are made to access memory hardware, response delays are noted and exploited to recover passwords, cryptographic keys, or other sensitive data. The current solution in which memory hardware is restricted to one program has been found to slow down computation. Therefore, US National Science Foundation (NSF) grantee researchers at the Massachusetts Institute of Technology (MIT) have developed a method that allows memory hardware to be shared without sacrificing security against memory timing side-channel attacks. Their approach increases computation speed by 12 percent compared to state-of-the-art solutions. The researchers focused on developing a solution to thwart contention attacks in which a malicious program tries to access memory hardware at the same time as another program. They developed a security scheme named DAGguise that uses a graph structure, known as a Directed Acyclic Graph (DAG), to process requests and submit the requests to the memory controller on a fixed schedule. Memory hardware can be shared among programs securely using the DAGguise structure. This article continues to discuss the DAGguise security scheme developed by MIT researchers to protect against memory timing side-channel attacks while allowing for dynamic traffic contention to achieve good performance. 

NSF reports "Is a Security Feature on the Way That Makes Computing Faster?"