I want to discuss the following scenario:

I use a cloud provider like amazon where every instance of the OS is a VM. The hypervisor launches the VMs as needed. So lets assume there are two VMs running, mine and attacker's on the same box at the same time.

Scenario 1: The attacker has root access to his VM. But does not have physical access to the box. What attack vectors do I worry about?

- Obviously any vulnerabilities in the my VM (OS and services it provides).  
  Lets assume this is out of scope for this discussion

- Vulnerabilities in the hypervisor that lets the attacker intercept all data to and from my VM. 
  Are there known vulnerabilities like this? 

- Any other scenarios like DMA attacks? 

Scenario 2: The attacker has physical access to the box. He manages the box and is able to install custom hardware or use the peripherals like USB, HDMI etc?

- I know that one can introspect the VM he wants to attack directly as software based attacks are easier. 
  Let's assume it's out of scope.

I want to discuss the following scenario:

I use a cloud provider like Amazon where every instance of the OS is a VM. The hypervisor launches the VMs as needed. So let's assume there are two VMs running, mine and attacker's on the same box at the same time.

Scenario 1: The attacker has root access to his VM. But does not have physical access to the box. What attack vectors do I worry about?

• Obviously any vulnerabilities in my VM (OS and services it provides).
  Let's assume this is out of scope for this discussion

• Vulnerabilities in the hypervisor that lets the attacker intercept all data to and from my VM. Are there known vulnerabilities like this?

• Any other scenarios like DMA attacks?

Scenario 2: The attacker has physical access to the box. He manages the box and is able to install custom hardware or use the peripherals like USB, HDMI etc?

• I know that one can introspect the VM he wants to attack directly as software-based attacks are easier. Let's assume it's out of scope.