Cloud computing security – identity and access management comparison for GCP and AWS
Cloud computing is the foundation of many digital transformation projects. Currently, enterprises see progressively more value in the ability to scale the cloud environment, being able to transfer part of the responsibility for reliability onto a cloud service provider as well as the accessibility of data for employees working remotely. However, the ease of access to data creates the need for rigorous identity management: specifying exactly who, in what situation and to what level can access the data stored in the cloud.
Each leading provider of IaaS cloud computing solutions: Google (GCP), Amazon (AWS) and Microsoft (Azure) assists in this process by granting access to Identity and Access Management tools. In spite of fairly similar scopes of functionality, the companies’ approaches to the design and defining policy of permissions and security, as well as user management and even the meaning of the same vocabulary can vary depending on the service provider.
In this article, you will find out about:
1. identity management in IAM systems,
2. different approaches to the structures used for access management within an organisation,
3. the differences between IaaS providers.
Identity management in IAM systems
Generally speaking, IAM systems are used to define and manage access. Online user identity management as well as a system of access control for IT infrastructure components are the most important elements of this system, allowing for them to grant the correct access level to the right infrastructure and under specific circumstances for individual persons.
User identification in the cloud
Every leading cloud computing services provider approaches the topic of user identification in their infrastructure differently.
GCP: the use of products offered by Google
For user identification, GCP uses other products offered by Google in. For example, we can gain access to the infrastructure through a Google account or an account from a G Suite domain. Google Cloud Identity can also be used for user management in a way resembling G Suite, but without access to office suite. Thanks to the separation of identity verification from IaaS, the user can access infrastructure belonging to different organisations through one account.
Fedareted identity management
Both providers, Amazon and Google allow for extension of mechanisms’ abilities through the incorporation of federated identity management based on SAML 2.0.
If properly set up, this functionality allows for the use of external services such as Microsoft Active Directory or an external provider such as Facebook, in which case AWS and GCP authentication processes are conducted based on user accounts from outside solutions and credentials is not stored using cloud services. This solution allows for the use of the same credentials for all configured services despite them being sourced from different organisations or different IaaS providers.
Additionally, this solution is often used alongside SSO (single sign-on) mechanisms, particularly for companies using IaaS services delivered by a number of providers, and is designed to help the user in navigating between them.
Adequate identity management in the cloud is the foundation of security and access management. Another crucial part of the process is defining different permission levels for each user, in which case IAM tools used can be quite different from one another.
Access management structures
Access management is usually implemented in order to allow users to perform operations needed for their assigned duties and limit user access level to areas necessary for each user.
The simplicity of AWS solutions
In the process of designing AWS structures, Amazon prioritised simplicity and granularity. In order to configure access policy, it is necessary first to specify the actions (mapped directly to API and AWS methods) and objects the policy applies to.
Each resource is given a unique ARN (Amazon Resource Name) ID when it is created.
During policy access configuration we target an object using its ARN. Due to the fact that each, even very specific objects, can be treated as resources, this solution allows for granting access with great accuracy, e.g. even to a single alarm in an AWS Cloud Watch or a Topic in SNS.
We can further specify access using wildcard characters.
The majority of ARN contain a path to a resource. Using an asterisk we can end the path on any chosen level, to refer to each object contained in a resource before the wildcard. For example, by granting a service access to arn:aws:iam::123456789012:user/*, we are giving it access to all user accounts.
Moreover, AWS can be used to create user accounts which do not have administrative access to infrastructure and are only granted access to API and CLI tools.
Access control hierarchy in GCP
Google has created a more complex access management solution based on resource hierarchy. In this solution, the Project is the foundation for creating an operational structure. All resources used in a Project have a basic level of access set by default. Thanks to this solution, e.g. App Engines are automatically granted access to all storage buckets created within a Project.
On account of the use of resource hierarchy, access can be granted not only to individual resources or users, but also to more general containers.
These containers can be:
- The previously mentioned Projects, containing resources related to a specific venture,
- Folders, which can store Projects as well as other Folders used to group resources,
- Organisations – the top of a resource hierarchy, containing all other objects.
How are permissions inherited in this structure?
For example by granting developer access on a Folder to specific a user, we are giving the user a similar role and the level of access that comes along with it for all Projects stored in that Folder and all resources contained within the Projects (as shown on the diagram below):
Although the structure of access in AWS might seem more intuitive, often subsequently to configuration, particularly of larger environments, we can encounter policies with permissions of wider reach than what is usually recommended, in order to facilitate management and speed up configuration. In these instances we can often encounter general definitions being applied to all of a certain type of resources (arn:aws:s3:::*) instead of specified resources.
On the other hand, in some instances GCP can lack the level of detail in defining access permissions, in which cases we have to circumvent the issue with more general permissions related to projects.
Different functions under the same name
In addition to different approaches to access management, when comparing GCP and AWS we have to overcome the issue of discrepancies in terminology used in the two solutions. Functions and objects under the same name work differently in GCP and AWS and often apply in different circumstances.
The most noticeable discrepancies exist in terminology related to Policy and Roles.
Policy in GCP and AWS
In AWS the Policy is a collection of individual permissions. We use it to group permissions granting access to a specific functionality. For example, one of the policies delivered and managed by Amazon “AmazonEC2ReadOnlyAccess” contain permissions to read status and configuration of services:
- EC2 Auto Scaling
as well as to read correlated services configurations e.g. EC2
Policies defined in this way can be attached to Roles, user groups or directly to a user or resource to determine their permissions. The closest equivalent to this solution in GCP is a Role.
Policy in Google cloud services describes the access to a specific resource (e.g. Project or Folder). Policy includes principles allocating Roles to specific users (so-called bindings). With a Policy attached to a Project, we can automatically check whether a user trying to access the data has a Role with a required access level. Additionally, Policy allows to add additional requirements the user has to meet to gain access (e.g. a specific timeframe to complete a task).
Role in GCP and AWS
The previously mentioned Role in GCP in many ways corresponds to Policy in AWS. It is a collection of individual permissions granting access to specific functionalities.
In AWS, a Role is a different solution- In a way, it can be seen as a user without a specified identity attached. The Role is not attributed to a specific person, and access to it is configured similarly to a regular user. Other users and resources can be granted temporary access to it in order to perform operations using the Role’s permissions. The solution is used mainly in identity federation, but also when apps need to download data or make changes in AWS structures.
The list of discrepancies between the two solutions could be much longer, in order to analyse them in more detail we recommend looking into GCP documentation comparing similarly named functionalities presented below:
|IAM role and instance profile
|Cloud IAM service account
|Managed in IAM. Identity federated to external identity
|Managed outside Cloud IAM. Identity federated to external identity management system.
|A document that explicitly lists permissions.
|A list of bindings. A binding binds a list of members to a role.
|Policy attached to an IAM user or group or a resource.
|Policy attached to resource.
|Deny by default.
|Deny by default.
|Predefined set of permissions
|Auditing IAM calls
Source: GCP Documentation license: Creative Commons 4.0
With the growing popularity of cloud computing services, there is a growing number of organisations using the services of more than one provider. In multi cloud environments functionalities such as identity federation as well as an in-depth knowledge of features specific to particular solutions created by providers is incredibly valuable.
Although they were designed in a completely different way, IAM systems in both GCP and AWS serve a similar purpose. Despite many differences between the tools, the number of similar functionalities in GCP and AWS has grown with their development over the years. Currently, both solutions allow for the implementation of similar security policies, even though they do so using different methods. Due to this, the choice between the providers is often determined by different aspects of their services.