I would like to run a batch job on-prem and access AWS resources in our account.
I think the recommendation is to create an IAM user, which will be a machine user. Since I don't have a way to assign a role to the on-prem machine, or federate with AWS identity, I'll create an access key and install it on the on-prem machine. What's the best way to link my machine user to a policy?
I can create an IAM policy which allows the required actions (reading AWS SSM Parameters).
But, how should I link the machine user to the policy? I'm setting up these users/policies with Pulumi. Some options I'm aware of:
I can create a role, but then I think the machine user would have to assume the role. (My understanding is that roles do not have immediate "membership", it's just that users have the ability to assume roles. Or, AWS infrastructure can be set up with a role, like an EC2 or an EKS cluster can act as a role. In the future I do plan to move this job's execution to AWS infrastructure, but for now that's not an option.) Is assuming a role easy, for example a aws sts CLI call that I could put in my batch job's startup script before calling the main binary?
Or I could just attach the policy directly to the machine user. Generally that's not recommended from what I've read: you should have a layer between users and policies so when users change what they're doing you have indirection. But in this case maybe that's fine.
Or finally I could create a user group, attach the policy to the group, and add the machine user as a member of the group. Is that layer of indirection useful / an appropriate use of groups, especially if I'm already managing these policies with IaC? Most documentation recommends roles for the user-to-policy indirection, so I'm hesitant to use groups that way. However, that seems to be the expected approach for human users (glad for feedback on that too).
"Is it better to use AWS IAM User Group, or IAM Role for users to assume?" says a group would help manage permissions for multiple users (but so does Pulumi and I only have 1 or 2 machine users); and a role would help separate access rights from long-lived credentials but it seems like rotating the machine user's access key would have that benefit too without the extra assume-role step.
Is there a policy action in IAM that will restrict the visibility of resources in the colnsole to only a tagged set? So for example can i use ec2:Describe with a StringEquals condition for the tag such that a group of users will only ever see the set of EC2 servers their project is using?
Ideally i dont want all users in this VPC to see all servers regardless if they are read only
Its not currently possible:
Note Currently, the Amazon EC2 ec2:Describe* API actions do not
support resource-level permissions, so you cannot control which
individual resources users can view in the console. Therefore, the *
wildcard is necessary in the Resource element of the above statement.
For more information about which ARNs you can use with which Amazon
EC2 API actions, see Supported Resource-Level Permissions for Amazon
EC2 API Actions.
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/iam-policies-ec2-console.html
On option you have, that might work is to use multiple, independent AWS accounts, so each user can get there own 'sandbox' to play in, but setup consolidate billing so that that aspect of the accounts are all shared by one master account.
The AWS official site reads role as a collection of permissions and group as a collection of users. But still they look the same to me. You attach policies to groups or roles, and then assign groups or roles to a user. What exactly are the differences between role and group?
Short answer for googlers: you can't assign role to user.
group is a bunch of users with the same policies
role is a preset of policies for service(s)
Users can asume roles according to AWS docs:
Assuming a Role
AWS Groups are the standard groups which you can consider as collection of several users and a user can belong to multiple groups.
AWS IAM Roles are all together different species; they operate like individual users except that they work mostly towards the impersonation style and perform communication with AWS API calls without specifying the credentials.
Given that IAM Roles are little different, I am emphasizing only that. There are several types of IAM Roles like EC2 IAM Roles, Lambda etc. If you consider, you can launch an EC2 instance with an EC2 IAM Role; hence forth any AWS API related communication wouldn't require any AWS Access Key or Secret key for authentication rather can call the APIs directly (however the long answer is - it uses STS and continuously recycles the credentials behind the scenes); the privileges or permissions of what it can do is determined by the IAM Policies attached to the IAM Role.
Lambda IAM Role works exactly the same, except that only Lambda function can use the Lambda IAM Role etc.
Users: End User (Think People).
Groups: A collection of users under one set of permissions (permission as policy). As per IAM standards we create groups with permissions and then assign user to that group.
Role: you create roles and assign them to AWS resource (AWS resource example can be a customer, supplier, contractor, employee, an EC2 instance, some external application outside AWS) but remember you can't assign role to user.
It’s not only users who will login, sometimes applications need access to AWS resources. For example, an EC2 instance might need to access one or more S3 buckets. Then, an IAM role needs to be created and attached to the EC2 instance. That role can be re-used by different EC2 instances.
Remember : Groups are for living. Roles are for non-living.
I think of an AWS Role as a kind of 'sudo', where each AWS Role can temporarily provide a very specific set of elevated privileges, but without needing the elevated credentials. I get the impression that like sudo, AWS Roles try to prevent privileged actions being used accidentally.
I'd be interested to hear if others agree with this analogy.
Please note that Groups are specific to local IAM users, which are not federated, and local IAM user logs do not show who has done the actions (i.e.., multiple people or applications could use the same long-term secret/access keys, and there is no record of which entity used them). If you must use local IAM users, you can place them into IAM Groups. Where this can be especially useful is to serve as a boundary -- you could place a deny policy on the group, restricting access to specific services or actions, and that deny policy will be applied to all users in the Group.
Conversely, roles can be federated, whereas local IAM users are not. You might create an on-premises AD group that serves as a member container, for example, and then the members of that AD group (and only they) can use the role that the AD group correlates to, with whatever allow or deny policies and/or permissions boundaries you've applied to the role. (Here is a link explaining the AWS ADFS federation.)
Importantly, roles allow for temporary session credentials (which is a best security practice), as their session tokens expire after a maximum of 12 hours. Equally importantly, roles do show in the logs which of the AD members with access to use the role actually did the action. You'll find this tacked to the end of the role ARN in the logs (e.g., a user ID). CloudTrail would be one of several services that indicate user activity. This is important from a logging standpoint.
Understanding IAM roles vs IAM groups (IAM indentities) is very important foundational concept . Its important to look at difference between IAM role and IAM user as essentially group is just a bunch of users performing similar functions (eg. group of developers, QA's etc.) Roles are not uniquely associated with one person (user), they can be assumed by user,resource or service who needs it to perform task at that point of time (session). Roles do not provide long-term credentials like password or access keys.
Best practice recommendation is to require workloads to use temporary credentials with IAM roles to access AWS
Please refer to link below for more clarity:
https://docs.aws.amazon.com/IAM/latest/UserGuide/id.html
I was confused all the time about the difference between these two functions.
In short,
Role is like a tag with all the preset policies that can attach on IAM users/groups or AWS services. IAM users share the same account with the account root user (Admin) but with assigned permissions by the root user to use AWS resources within that account.
Therefore, IAM users can directly interact with AWS services; whereas IAM roles cannot make direct requests to AWS services, they are meant to be assumed by authorised entities like an IAM user or an instance. https://aws.amazon.com/iam/faqs/
I had a hard time deciphering the spirit of the given answers..
Here's what I've found:
Groups:
Intended to represent human users created within IAM who need identical policies.
Ex. Dev 1 - Dev 8 are all developers, and all need access to create dev servers.
This is similar to traditional desktop users/groups, but for HUMAN users only.
Roles:
Roles rotate automatic credentials, meaning password input isn't needed for accessing policies.
This makes it good for two things:
Giving permissions to non-humans, such as services / applications.
Ex. EC2 of type A needs access to S3 of type B.
Giving permissions to federated / outside users & groups.
Ex. Contractor A # Outside Company A needs access to your Server A.
Authentication of users & groups are handled by some service, like Azure AD.
Authorizations are then mapped to your IAM role(s), NOT users or groups.
Note: I've used Jumpcloud's Article & AWS's Documentation to gather this information. The terms "Group", "Role", and "User" become overloaded in context to SSO+IdP, and IAM.
Here's an image showing how they map roles. !Need 10 Reputation :(
Aside: There is a way of assigning Roles to normal IAM Users & Groups, but it appears to be bad practice.
Hopefully this provides clarity to the answers above.
Only one IAM Role can be assumed at a time! And there are several
situations which fits exactly this kind of permission.
Read the faq about: How many IAM roles can I assume?
The underlaying tool in use is "Permission" in both of the use cases namely: Group and IAM Role.
Group or IAM Role --> Has Policy --> Policy defines permisions --> Permissions are assigned to a Group or IAM Role.
I would like to write a policy for a new IAM user so that he can have a full access to AWS Services such as EC2, RDS, Cloud Front, S3, etc. However he should be only able to view (describe) and manage the instances/services that he launches. He can't see other existing/future instances/services created by other users.
This case happens in a company that wants to outsource some projects to an outsourced company. So the new IAM user is for giving access to the outsourced company to setup the staging & production environments in AWS that belongs to the company. How can I achieve this?
Thanks.
Best Regards,
Mark
AFAICT there are no conditions on creator.
One obvious alternative is having a separated dependent and linked account with consolidated billing. You can find some detail in the third scenario of this paper
https://media.amazonwebservices.com/AWS_Setting_Up_Multiuser_Environments_Education.pdf
[...] an administrator creates separate AWS accounts for each user who
needs a new AWS account. These accounts can optionally be linked
together and a single AWS account can be designated as the paying
account using consolidated billing, which provides a single bill for
multiple AWS accounts. The administrator then creates an IAM user in
each AWS account and applies an access control policy to each user.
Users are given access to the IAM user within their AWS account, but
do not have access to the root credentials of the AWS account.
Users can log into the AWS Management Console with their IAM
credentials and then they can launch and access different AWS
services, subject to the access control policies applied to their
account. Users have direct control over the access credentials for
their resources and they can also share these resources with other
users as necessary.
I am in the early stages of writing an AWS app for our users that will run our research algorithms using their AWS resources. For example, our code will need to spin up EC2 instances running our 'worker' app, access RDS databases, and create access SQS queues. The AWS Java SDK examples (we are writing this in Java) use a AwsCredentials.properties file to store the Access Key ID and Secret Access Key, which is fine for examples, but obviously not acceptable for our users, who are would be in essence giving us access to all their resources. What is a clean way to go about running our system on their behalf? I discovered AWS Identity and Access Management (IAM) which seems to be for this purpose (I haven't got my head around it yet), esp. Cross-account access between AWS accounts. This post makes it sound straightforward:
Use the amazon IAM service to create a set of keys that only has
permission to perform the tasks that you require for your script.
http://aws.amazon.com/iam/
However, other posts (e.g., Within IAM, can I restrict a group of users to access/launch/terminate only certain EC2 AMIs or instances?) suggest there are limitations to using IAM with EC2 in particular.
Any advice would be really helpful!
The key limitation with regards to RDS and EC2 is that while you can restrict access to certain API actions there are no resource level constraints. For example with an IAM S3 policy you can restrict a user to only being able to perform certain actions on certain buckets. You can write a policy for EC2 that says that user is allowed to stop instances, but not one that says you can only stop certain instances.
Another option is for them to provide you with temporary credentials via the Security Token Service. Another variant on that is to use the new IAM roles service. With this an instance has a set of policies associated with it. You don't need to provide an AwsCredentials.proprties file because the SDK can fetch credentials from the metadata service.
Finally one last option might be consolidated billing. If the reason you are using their AWS resources is just because of the billing, then setup a new account which is billed from their account. The accounts are isolated from each other so you can't for example delete their instances by accident. Equally you can't access their RDS snapshots and things like that (access to an RDS instance via mysql (as opposed to the AWS api) would depend on the instance's security group). You can of course combine this with the previous options - they could provide you with credentials that only allow you to perform certain actions within that isolated account.