Abstract
Error detection and correction approaches are often used in modern microprocessors to detect and mitigate soft errors. However, when such errors must be handled in software, transitioning to the error handler may require parts of the processor state to be hidden and/or flushed to prevent information leakage. This paper presents a high level introduction to some issues in this area. For example, it discusses when security checks may be performed and what actions may be taken to help preserve isolation when the error handler is at a trust level other than the executing code.