Only allow EC2 instance to access static website on S3 - amazon-web-services

Question

I have a static website hosted on S3, I have set all files to be public.
Also, I have an EC2 instance with nginx that acts as a reverse proxy and can access the static website, so S3 plays the role of the origin.
What I would like to do now is set all files on S3 to be private, so that the website can only be accessed by traffic coming from the nginx (EC2).
So far I have tried the following. I have created and attached a new policy role to the EC2 instance with
Policies Granting Permission: AmazonS3ReadOnlyAccess
And have rebooted the EC2 instance.
I then created a policy in my S3 bucket console > Permissions > Bucket Policy
{
"Version": "xxxxx",
"Id": "xxxxxxx",
"Statement": [
{
"Sid": "xxxxxxx",
"Effect": "Allow",
"Principal": {
"AWS": "arn:aws:iam::XXXXXXXXXX:role/MyROLE"
},
"Action": "s3:GetObject",
"Resource": "arn:aws:s3:::XXX-bucket/*"
}
]
}
As principal I have set the ARN I got when I created the role for the EC2 instance.
"Principal": {
"AWS": "arn:aws:iam::XXXXXXXXXX:role/MyROLE"
},
However, this does not work, any help is appreciated.

Answer 1

If the Amazon EC2 instance with nginx is merely making generic web requests to Amazon S3, then the question becomes how to identify requests coming from nginx as 'permitted', while rejecting all other requests.
One method is to use a VPC Endpoint for S3, which allows direct communication from a VPC to Amazon S3 (rather than going out an Internet Gateway).
A bucket policy can then restrict access to the bucket such that it can only be accessed via that endpoint.
Here is a bucket policy from Example Bucket Policies for VPC Endpoints for Amazon S3:
The following is an example of an S3 bucket policy that allows access to a specific bucket, examplebucket, only from the VPC endpoint with the ID vpce-1a2b3c4d. The policy uses the aws:sourceVpce condition key to restrict access to the specified VPC endpoint.
{
"Version": "2012-10-17",
"Id": "Policy",
"Statement": [
{
"Sid": "Access-to-specific-VPCE-only",
"Action": "s3:*",
"Effect": "Allow",
"Resource": ["arn:aws:s3:::examplebucket",
"arn:aws:s3:::examplebucket/*"],
"Condition": {
"StringEquals": {
"aws:sourceVpce": "vpce-1a2b3c4d"
}
},
"Principal": "*"
}
]
}
So, the complete design would be:
Object ACL: Private only (remove any current public permissions)
Bucket Policy: As above
IAM Role: Not needed
Route Table configured for VPC Endpoint

Answer 2

Permissions in Amazon S3 can be granted in several ways:
Directly on an object (known as an Access Control List or ACL)
Via a Bucket Policy (which applies to the whole bucket, or a directory)
To an IAM User/Group/Role
If any of the above grant access, then the object can be accessed publicly.
Your scenario requires the following configuration:
The ACL on each object should not permit public access
There should be no Bucket Policy
You should assign permissions in the Policy attached to the IAM Role
Whenever you have permissions relating to a User/Group/Role, it is better to assign the permission in IAM rather than on the Bucket. Use Bucket Policies for general access to all users.
The policy on the Role would be:
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "AllowBucketAccess",
"Effect": "Allow",
"Action": [
"s3:GetObject"
],
"Resource": [
"arn:aws:s3:::my-bucket/*"
]
}
]
}
This policy is directly applied to the IAM Role, so there is no need for a principal field.
Please note that this policy only allows GetObject -- it does not permit listing of buckets, uploading objects, etc.
You also mention that "I have set all files to be public". If you did this by making each individual object publicly readable, then anyone will still be able to access the objects. There are two ways to prevent this -- either remove the permissions from each object, or create a Bucket Policy with a Deny statement that stops access, but still permits the Role to get access.
That's starting to get a bit tricky and hard to maintain, so I'd recommend removing the permissions from each object. This can be done via the management console by editing the permissions on each object, or by using the AWS Command-Line Interface (CLI) with a command like:
aws s3 cp s3://my-bucket s3://my-bucket --recursive --acl private
This copies the files in-place but changes the access settings.
(I'm not 100% sure whether to use --acl private or --acl bucket-owner-full-control, so play around a bit.)