A Virtualized Harvard Architectural Approach to Protect Kernel Code

Most modern computers which implement the von Neumann architecture have the risk of being attacked by modifying kernel code, because code is put in the same address space as the data. In this paper, the virtualization of Harvard memory architecture is implemented as a patch to the current operating system to separate the code fetch and the data operation on the kernel code, and thus solve the problem at its root. The implementation of our virtualization is described in detail, its effectiveness to resist attacks is evaluated and the impact on the current operating system is fully analyzed. The experiment results show our approach is effective, has no side effect on the normal functions of the operating system and incurs very little overhead.