Security and Separation with different Cloud service models

Security and Separation with Cloud service models

The technical controls required to provide separation will vary depending on the service model employed. We’ll consider these for each service type in turn.

Infrastructure as a Service (IaaS)

IaaS products generally provide compute, networking and storage services. User separation needs to be enforced in all of these elements.

IaaS: compute separation

Within the compute environment, separation between users is typically enforced by a hypervisor (though in some circumstances it may also be achieved by allocating physical hardware to consumers).

The strength of separation usually depends on the virtualisation technology in use. Using hardware virtualisation and assured virtualisation products should provide higher confidence in the separation within the compute environment. Obviously, this depends on the correct configuration and operation of the product.

The administration tools supporting the virtualisation product should be secured too, as they are fundamental to the security provided by the product.

IaaS: network separation

It is important to understand the network separation model of an IaaS offering since users will normally be constructing their own virtual networks on multi-tenanted network infrastructure.

A number of technologies could be used by the service provider to enforce network separation. These include the use of virtual LANs (VLANs), virtual routing and forwarding (VRF) technologies, or virtual networking capabilities within the compute environment.

If you cannot gain sufficient confidence in the separation provided by an IaaS service within the networking layer, you may enhance your confidentiality protection with your own encryption. This is described in our IaaS configuration guide.

You may also need to consider whether the service protects or reserves your share of network resources, so that an attack (such as a DDoS) on another user does not affect your service provision.

IaaS: storage

In an IaaS model, users may have direct control over some portion of a multi-tenanted storage environment. Providing users with this direct control makes it possible for a malicious or compromised user to attack the storage components of the service in order to gain access to another user’s data. Effective separation within the storage of the service will help tackle this problem.

IaaS users can reduce their reliance on the cloud service provider’s storage separation by encrypting their own data. This presents its own challenges, and will only be effective if the encryption keys for the data can be stored in such a way that they would not be accessible to an attacker who has access to the storage.

Platform as a Service (PaaS)

PaaS offerings can provide users with rich and complex interfaces. They cover a wide range of implementation technologies, and are likely to be at different levels of security maturity.

Depending on the technologies involved, it may be difficult, time-consuming and expensive to gain high levels of assurance in the separation provided by a PaaS offering.

Applications wanting more assurance in the separation provided by a PaaS offering may instead build upon an IaaS offering which has sufficient assurance in such a way that separation will be enforced by the underlying IaaS. Alternatively, it may be appropriate to run the PaaS solution within a private or community cloud service.

Two common PaaS approaches, with different associated risks, are described below.

Shared application hosting

A common PaaS model, particularly for the delivery of web application hosting, involves the hosting of users’ applications on top of a shared operating system. In this model, the operating system and application host (eg the web server and the scripting host or runtime) are responsible for preventing users from affecting each other.

If attackers can legitimately run an application on the same host, they have access to a large attack surface to attempt to escalate privileges and gain unauthorised access.

Managed host services

Another common PaaS model is the provision of managed operating system services. In this model, the user has a dedicated physical or virtual machine, but rather than manage the operating system themselves, they purchase this as a service from the service provider.

The risks in this model are very similar to that of an IaaS service, since the separation enforcement is likely to be based on the same underlying technology (typically a hypervisor).

There are two key additional risks (compared to an equivalent IaaS service) to bear in mind when considering the suitability of this model:

  • The service provider’s administrators will have privileged access to the operating system. This means it is easier for them to access your data than if they only had access to disk images.
  • In providing the management service, additional connections between your machines and the management infrastructure are likely. This infrastructure is an additional attack vector compared to the simple IaaS case.

Software as a Service (SaaS)

In SaaS offerings the separation between users is often enforced by software controls running within a single instance of an application. The strength of the separation is dependent on the application architecture and implementation.

The underlying platform and infrastructure is not usually relied upon to enforce separation, which makes it difficult to gain high degrees of confidence in the strength of separation. In SaaS offerings it may be difficult to understand where and how your data is protected.

Some SaaS offerings may be dedicated to a single consumer, leveraging the controls of an IaaS or PaaS solution to give users confidence in the separation of their data.

In SaaS offerings the service provider will typically rely on application level controls rather than controls in the infrastructure or platform – meaning that if a component in the service is compromised then the data of many users may be visible to that component.

If you need more assurance in the separation provided by an SaaS offering you may want to build upon an IaaS or PaaS offering which has sufficient assurance in the underlying separation controls. Alternatively, it may be appropriate to run the SaaS solution within a private or community cloud service.

Risks associated with service models

The following table summarises the main risks associated with each of the service models.

Service model Associated risks
Infrastructure as a Service (IaaS) Offerings implemented using hardware virtualisation and leading virtualisation products can provide a good level of separation between workloads and data in community and public cloud platforms.

However, like all complex software, IaaS offerings will never be free from vulnerabilities and the risks that these bring.

IaaS services also have a much greater burden on the user to configure and operate well.

Platform as a Service (PaaS) PaaS offerings tend to have a larger attack surface than IaaS offerings since the separation between users is normally provided in higher level software rather than by a hypervisor. Community cloud PaaS offerings may provide some additional comfort for users where an acceptable use policy is in place that has been designed to reduce the risk of malicious workloads.

PaaS technologies are evolving rapidly and you should regularly verify that your platform choice meets your business and security needs.

Software as a Service (SaaS) SaaS offerings tend to implement separation at a higher level than both IaaS and PaaS, meaning the potential attack surface for a would-be attacker is much greater.

Unless architected well these services will often present a potentially higher risk than deploying software packages for a dedicated user within an IaaS or PaaS service.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.