HyperMI: A Privilege-level Secure Execution Environment

Table of Contents

• Introduction
  • The modified Host OS kernel
  • The HyperMI World
  • The Switch Gate
• Compilation and Installation
  • Compilation and Installation Steps
  • Modified Files
• Acknowledgement

Introduction

HyperMI is a secure and in-kernel isolated execution environment to protect Virtual Machines (VMs) against the compromised hypervisor. Currently, we designed and implemented HyperMI on the KVM-based cloud computing environment.

HyperMI, as shown in Figure 1, consists of three parts: the modified host OS kernel (not trusted), the HyperMI World and the Switch Gate.

The Modified Host OS Kernel

Since the hypervisor/Host OS is not trusted by us, access to some critical data structures must be restricted. The following modifications were made to the Host OS Kernel by HyperMI

1. Virtual Machines Control Structure (VMCS) and Extended Page Table (EPT) are removed from the original kernel.
2. Functions which are relevant to Control Registers (CR0, CR3, CR4), VMCS, EPT, DMA are hooked.

The HyperMI World

The HyperMI World is the secure and in-kernel isolated execution environment to host important data structures and security tools. HyperMI has full capability of monitoring the Hypervisor and protecting critical data structures. Currently, the HyperMI world hosts the following data structures and security tools:

1. Important Data Structures:
   • Virtual Machines Control Structure (VMCS)
   • Extended Page Table (EPT)
2. Security Tools:
   • Verification Program for Control Registers and DMA Mapping
   • Physical Page Tag Table (PPTT)

Verification Program for Control Registers and DMA Mapping is used to prevent the compromised hypervisor/Host OS kernel from subverting our HyperMI World.
EPT behaviors are controlled by the three control bits in PPTT: PFN(Physical Page Frame Number), OwnerID and Shared ID. With the help of PPTT, the way to subvert the isolation of different VMs through dedicated EPT entries is hindered.

The Switch Gate

The Switch Gate is the only interface between the Normal World (the compromised hypervisor/Host OS Kernel) and the HyperMI world. Figure 2 shows the detail of this Switch Gate.

The Switch Gate may be similar with the Switch Gate proposed by SKEE at the first look. But, there are several differences between them:

1. HyperMI implemented the Switch Gate in x86_64 environment, while SKEE just implemented their work based on ARM v7 and ARM v8.
2. HyperMI implemented the Switch Gate by using a new kernel page table hosted in the HyperMI world, the access to the original kernel page table are restricted and redirected to the kernel page table in the HyperMI world. This approach is totally different from the approach adopted by SKEE.

Compilation and Installation

Compilation and Installation Steps

The compilation and installation process of HyperMI contains two step:

1. Compile and install the modified linux kernel

   git clone https://github.com/DCSec/HyperMI.git 
   cd HyperMI 
   cp -r src/linux-4.4.1 /usr/src
   cd /usr/src/linux-4.4.1
   make oldcofigure && make modules && make modules_install&& make install 

2. Compile and install SecureVM

   cd home/to/HyperMI
   cd src/SecureVM
   make && insmod secureVM.ko

Modified Files

HyperMI requires a modified kernel to provide the secure and in-kernel isolation execution environment. The modified source code files are listed as following:

    fault.c 
    ksm.h 
    kvm_host.h 
    mmu.c 
    mmu.h 
    mmu_audit.c 
    pgtable_64.h 
    secureVM.h 
    types.h 
    virtio_balloon.c 
    vmscan.c 
    vmx.c 
    vmx.h 
    vmx_header_modify.h 
    vmx_modify.h

Acknowledgement

• This work has been pushlished in The 18th IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom 2019).