A loop statement execution method and device for side channel security protection

Abstract

The invention discloses a loop statement execution method and device for side channel security protection. The loop statement execution method for side channel security protection of the present invention includes pre-identifying the source code of the target program or the loop body in the target code, which is for each segment Insert an extended loop instruction based on the extensible instruction set architecture coding specification before the loop body to obtain the target program using the extended loop instruction; when executing the target program using the extended loop instruction, if an extended loop instruction is encountered, the extended loop instruction will be executed based on the extended loop instruction. The instruction executes the loop body after the extended loop instruction deterministically. The invention can resist branch prediction attack, not only guarantees security, but also eliminates the huge performance loss brought by the existing defense measures to the loop, and also controls the increase of the code amount to only one per loop compared with loop unrolling, There are few modifications to existing programs.

Description

technical field [0001] The invention belongs to the field of processor architecture security, and in particular relates to a loop statement execution method and device for side channel security protection. Background technique [0002] Branch prediction technology is a key technology used by modern processors to improve program execution efficiency. It greatly reduces the pipeline delay caused by branch instructions. High-accuracy branch prediction can greatly improve processor performance. However, in recent years, the Spectre series of vulnerabilities have been exposed, showing that there are serious security problems in the design of typical modern processor architectures, posing a great security threat to processor users. Spectre exploits misleading training on branch prediction techniques, allowing processors to speculatively execute code sequences crafted by attackers, unauthorized access to data protected by various processor and operating system security mechanisms, ...

AI Technical Summary

Problems solved by technology

This type of defensive measure is equivalent to suspending the pipeline after the processor executes a branch instruction, and then executes it after the branch instruction is determined. In this way, the processor is equivalent to not using branch prediction technology, which leads to a great decrease in program execution efficiency and sacrifices the branch Huge performance gains from predictive techniques come at great cost

The program contains a lot of loop statements, which is also a necessary condition for attackers to use to construct "ghost" attacks. Existing measures ensure processor security by completely disabling branch prediction, but when executing branch instructions, especially branch instructions in loops can incur huge overhead when

Images