One of the top cloud computing threats involves one of its core enabling technologies: virtualization. In virtual environments, there are multiple attack avenues, but this tip will look at ones that are most likely to be experienced by users of the technology:
- Prebuilt virtual machines/virtual appliances containing malicious code (Trojaned)
- Improperly configured virtual firewalls or networking
- Improperly configured hypervisor
- Data leakage through offline images
When building security defenses against these
Trojaned virtual machines/virtual appliances
The presence of untrusted virtual machines or Trojaned virtual appliances in the environment should be the first virtualization security risk you address. The untrusted virtual machine will manifest itself in public clouds (i.e., multitenant), and is a bad guy bringing up a malicious system that will attempt to identify “proximity” related vulnerabilities. The threats are created because the VM is either running on the same hypervisor or within the same cloud, and the cloud provider has created some level of “trust” between the virtual machines that the consumer is not aware of. If those vulnerabilities exist, the likelihood of exploit increases significantly.
Malicious virtual appliances (an appliance in this sense is anything that is “pre-packaged” for you to just download and run as a VM) would be a threat in public or private cloud environments. Since you install/use these appliances, there’s an element of trust you have given them. The malicious system would then attempt to find vulnerabilities through its “trust” and exploit them. Now an attacker would have a compromised machine in your environment, thus the intrusion succeeded. Recently, Amazon notified its Elastic Compute Cloud customers that it had identified compromised Amazon Machine Images (AMI) in its community set of AMIs, which are combinations (i.e., stacks) of software created to help users deploy servers quickly in EC2. The notification reminded users about the danger of compromised AMIs. Amazon realized that a compromised “appliance” or “build image” provides a trusted foothold in obtaining critical information, such as credentials for further exploit.
The keys to preventing these threats are to only use verified and tested appliances/images, and have assurance that your cloud provider has properly configured hypervisor and networking configurations that do not create unintended proximity trust. Further, keep in mind the nature of physical isolation/separation in the virtualized environment and configure systems accordingly.
Improperly configured virtual firewalls or networking
Traditionally, the networking team handled the configuration of firewalls and network equipment. They understand the intricacies and security implications of VLANs, tagging, routing, stateful connections, how inbound vs. outbound apply to interfaces, etc. This may or may not be true for many host administrators. In a virtual environment, many of the host administrators are now configuring and managing these network security devices. If these devices are not configured correctly, you can have traffic meant for one VM being seen or delivered to another VM or an outside entity.
While tools exist to do this right, it is primarily a people “issue,” in that administrators are being asked to manage security devices they do not understand. Overlook this, and you can have your underbelly exposed.
The key to minimizing this virtualization security risk is to have the network team handle the networking in the virtual environment, even though the virtual networking devices are not physical pieces of networking equipment. If you don’t have that luxury, then training the host administrator in the security aspects of network configurations is the next best mitigation.
Improperly configured hypervisor
The security of the environment is linked to the security of the hypervisor; any unauthorized access to the hypervisor compromises the environment. The main threat here is a lack of controls to limit who can gain access, and once in, what access they have. These are pretty straightforward threats, in that allowing unrestricted access to hypervisors, especially one that can be reached from an untrusted network (e.g., the Internet) increases the threat of attack. Secondly, if you do not restrict what legitimate users can access once they authenticate, you open yourself up to attack.
The first vulnerability is easy to deal with by restricting what systems can get access to the management functions (GUI, API, login, etc.). The second is not quite as easy, because you need a robust access control mechanism to allow the access/management of the users VMs (workloads), but not to the host. Further, you likely want to allow certain operations on the host to some users, but not all operations. The underlying access control mechanism must be able to support this. Since many cloud providers (e.g., Amazon) are building hypervisors almost from the ground up, getting the necessary access control mechanisms is difficult. .
The key to minimizing this threat is the use of granular role-based access control mechanisms to the hypervisor and management applications.
Data leakage through offline images
When guest images are suspended, unlike physical systems that would need physical access to pull data out of memory, the memory “footprint” now is in a file, and for all intents, searchable. Take an application securing Social Security or credit card numbers: The VM may be solid and secure, however, when the system is suspended, any information put in memory is likely not protected and assumed to be volatile. The problem arises when the image suspends and writes that memory to disk. Also, with migration, information states that would never have existed before now exist. We need to protect against that.
The nice thing about virtual security is that if you have a good security program in general, you will likely have a secure virtual environment as well. The fact is that if you apply old system administration practices (as they relate to security) to virtual environments, you will be well on the way to minimizing virtualization security risks and a secure environment.
About the author:
Philip Cox is Director of Security and Compliance, for SystemExperts Corporation, a consulting firm that specializes in system security and management. He is a well-known authority in the areas of system integration and security. He serves on the Trusted Cloud Initiative Architecture workgroup, as well as the PCI Virtualization and Scoping SIGs. Phil frequently writes and lectures on issues dealing with heterogeneous system integration and compliance (PCI-DSS and ISO). He is the lead author of Windows 2000 Security Handbook Second Edition(Osborne McGraw-Hill) and contributing author for Windows NT/2000 Network Security (Macmillan Technical Publishing) and CIW Security Professional Certification Bible (Wiley).
This was first published in April 2011