August 1st, 2022: Post updated to clarify how GuardDuty Malware Protection works with KMS keys.
With Amazon GuardDuty, you can monitor your AWS accounts and workloads to detect malicious activity. Today, we are adding to GuardDuty the capability to detect malware. Malware is malicious software that is used to compromise workloads, repurpose resources, or gain unauthorized access to data. When you have GuardDuty Malware Protection enabled, a malware scan is initiated when GuardDuty detects that one of your EC2 instances or container workloads running on EC2 is doing something suspicious. For example, a malware scan is triggered when an EC2 instance is communicating with a command-and-control server that is known to be malicious or is performing denial of service (DoS) or brute-force attacks against other EC2 instances.
GuardDuty supports many file system types and scans file formats known to be used to spread or contain malware, including Windows and Linux executables, PDF files, archives, binaries, scripts, installers, email databases, and plain emails.
When potential malware is identified, actionable security findings are generated with information such as the threat and file name, the file path, the EC2 instance ID, resource tags and, in the case of containers, the container ID and the container image used. GuardDuty supports container workloads running on EC2, including customer-managed Kubernetes clusters or individual Docker containers. If the container is managed by Amazon Elastic Kubernetes Service (Amazon EKS) or Amazon Elastic Container Service (Amazon ECS), the findings also include the cluster name and the task or pod ID so application and security teams can quickly find the affected container resources.
As with all other GuardDuty findings, malware detections are sent to the GuardDuty console, pushed through Amazon EventBridge, routed to AWS Security Hub, and made available in Amazon Detective for incident investigation.
How GuardDuty Malware Protection Works
When you enable malware protection, you set up an AWS Identity and Access Management (IAM) service-linked role that grants GuardDuty permissions to perform malware scans. When a malware scan is initiated for an EC2 instance, GuardDuty Malware Protection uses those permissions to take a snapshot of the attached Amazon Elastic Block Store (Amazon EBS) volumes that are less than 1 TB in size and then restore the EBS volumes in an AWS service account in the same AWS Region to scan them for malware. You can use tagging to include or exclude EC2 instances from those permissions and from scanning. In this way, you don’t need to deploy security software or agents to monitor for malware, and scanning the volumes doesn’t impact running workloads. The EBS volumes in the service account and the snapshots in your account are deleted after the scan. Optionally, you can preserve the snapshots when malware is detected.
The service-linked role grants GuardDuty access to AWS Key Management Service (AWS KMS) keys used to encrypt EBS volumes. If the EBS volumes attached to a potentially compromised EC2 instance are encrypted with a customer-managed key, GuardDuty Malware Protection uses the same key to encrypt the replica EBS volumes as well. The KMS keys from which GuardDuty issues grants to the service account cannot be invoked from any context except the Amazon EBS service. Further, the grants are retired after the scan completes. If the volumes are not encrypted, GuardDuty uses its own key to encrypt the replica EBS volumes to ensure that these volumes created by GuardDuty are always encrypted. Volumes encrypted with EBS-managed keys are not supported.
Security in cloud is a shared responsibility between you and AWS. As a guardrail, the service-linked role used by GuardDuty Malware Protection cannot perform any operation on your resources (such as EBS snapshots and volumes, EC2 instances, and KMS keys) if it has the GuardDutyExcluded tag. Once you mark your snapshots with GuardDutyExcluded set to true, the GuardDuty service won’t be able to access these snapshots. The GuardDutyExcluded tag supersedes any inclusion tag. Permissions also restrict how GuardDuty can modify your snapshot so that they cannot be made public while shared with the GuardDuty service account.
Enabling Malware Protection for an AWS Account
If you’re not using GuardDuty yet, Malware Protection is enabled by default when you activate GuardDuty for your account. Because I am already using GuardDuty, I need to enable Malware Protection from the console. If you’re using AWS Organizations, your delegated administrator accounts can enable this for existing member accounts and configure if new AWS accounts in the organization should be automatically enrolled.
In the GuardDuty console, I choose Malware Protection under Settings in the navigation pane. There, I choose Enable and then Enable Malware Protection.
Snapshots are automatically deleted after they are scanned. In General settings, I have the option to retain in my AWS account the snapshots where malware is detected and have them available for further analysis.
In Scan options, I can configure a list of inclusion tags, so that only EC2 instances with those tags are scanned, or exclusion tags, so that EC2 instances with tags in the list are skipped.
Testing Malware Protection GuardDuty Findings
To generate several Amazon GuardDuty findings, including the new Malware Protection findings, I clone the Amazon GuardDuty Tester repo:
$ git clone https://github.com/awslabs/amazon-guardduty-tester
First, I create an AWS CloudFormation stack using the guardduty-tester.template file. When the stack is ready, I follow the instructions to configure my SSH client to log in to the tester instance through the bastion host. Then, I connect to the tester instance:
From the tester instance, I start the guardduty_tester.sh script to generate the findings:
$ ./guardduty_tester.sh
***********************************************************************
* Test #1 - Internal port scanning *
* This simulates internal reconaissance by an internal actor or an *
* external actor after an initial compromise. This is considered a *
* low priority finding for GuardDuty because its not a clear indicator*
* of malicious intent on its own. *
***********************************************************************
Starting Nmap 6.40 ( http://nmap.org ) at 2022-05-19 09:36 UTC
Nmap scan report for ip-172-16-0-20.us-west-2.compute.internal (172.16.0.20)
Host is up (0.00032s latency).
Not shown: 997 filtered ports
PORT STATE SERVICE
22/tcp open ssh
80/tcp closed http
5050/tcp closed mmcc
MAC Address: 06:25:CB:F4:E0:51 (Unknown)
Nmap done: 1 IP address (1 host up) scanned in 4.96 seconds
-----------------------------------------------------------------------
***********************************************************************
* Test #2 - SSH Brute Force with Compromised Keys *
* This simulates an SSH brute force attack on an SSH port that we *
* can access from this instance. It uses (phony) compromised keys in *
* many subsequent attempts to see if one works. This is a common *
* techique where the bad actors will harvest keys from the web in *
* places like source code repositories where people accidentally leave*
* keys and credentials (This attempt will not actually succeed in *
* obtaining access to the target linux instance in this subnet) *
***********************************************************************
2022-05-19 09:36:29 START
2022-05-19 09:36:29 Crowbar v0.4.3-dev
2022-05-19 09:36:29 Trying 172.16.0.20:22
2022-05-19 09:36:33 STOP
2022-05-19 09:36:33 No results found...
2022-05-19 09:36:33 START
2022-05-19 09:36:33 Crowbar v0.4.3-dev
2022-05-19 09:36:33 Trying 172.16.0.20:22
2022-05-19 09:36:37 STOP
2022-05-19 09:36:37 No results found...
2022-05-19 09:36:37 START
2022-05-19 09:36:37 Crowbar v0.4.3-dev
2022-05-19 09:36:37 Trying 172.16.0.20:22
2022-05-19 09:36:41 STOP
2022-05-19 09:36:41 No results found...
2022-05-19 09:36:41 START
2022-05-19 09:36:41 Crowbar v0.4.3-dev
2022-05-19 09:36:41 Trying 172.16.0.20:22
2022-05-19 09:36:45 STOP
2022-05-19 09:36:45 No results found...
2022-05-19 09:36:45 START
2022-05-19 09:36:45 Crowbar v0.4.3-dev
2022-05-19 09:36:45 Trying 172.16.0.20:22
2022-05-19 09:36:48 STOP
2022-05-19 09:36:48 No results found...
2022-05-19 09:36:49 START
2022-05-19 09:36:49 Crowbar v0.4.3-dev
2022-05-19 09:36:49 Trying 172.16.0.20:22
2022-05-19 09:36:52 STOP
2022-05-19 09:36:52 No results found...
2022-05-19 09:36:52 START
2022-05-19 09:36:52 Crowbar v0.4.3-dev
2022-05-19 09:36:52 Trying 172.16.0.20:22
2022-05-19 09:36:56 STOP
2022-05-19 09:36:56 No results found...
2022-05-19 09:36:56 START
2022-05-19 09:36:56 Crowbar v0.4.3-dev
2022-05-19 09:36:56 Trying 172.16.0.20:22
2022-05-19 09:37:00 STOP
2022-05-19 09:37:00 No results found...
2022-05-19 09:37:00 START
2022-05-19 09:37:00 Crowbar v0.4.3-dev
2022-05-19 09:37:00 Trying 172.16.0.20:22
2022-05-19 09:37:04 STOP
2022-05-19 09:37:04 No results found...
2022-05-19 09:37:04 START
2022-05-19 09:37:04 Crowbar v0.4.3-dev
2022-05-19 09:37:04 Trying 172.16.0.20:22
2022-05-19 09:37:08 STOP
2022-05-19 09:37:08 No results found...
2022-05-19 09:37:08 START
2022-05-19 09:37:08 Crowbar v0.4.3-dev
2022-05-19 09:37:08 Trying 172.16.0.20:22
2022-05-19 09:37:12 STOP
2022-05-19 09:37:12 No results found...
2022-05-19 09:37:12 START
2022-05-19 09:37:12 Crowbar v0.4.3-dev
2022-05-19 09:37:12 Trying 172.16.0.20:22
2022-05-19 09:37:16 STOP
2022-05-19 09:37:16 No results found...
2022-05-19 09:37:16 START
2022-05-19 09:37:16 Crowbar v0.4.3-dev
2022-05-19 09:37:16 Trying 172.16.0.20:22
2022-05-19 09:37:20 STOP
2022-05-19 09:37:20 No results found...
2022-05-19 09:37:20 START
2022-05-19 09:37:20 Crowbar v0.4.3-dev
2022-05-19 09:37:20 Trying 172.16.0.20:22
2022-05-19 09:37:23 STOP
2022-05-19 09:37:23 No results found...
2022-05-19 09:37:23 START
2022-05-19 09:37:23 Crowbar v0.4.3-dev
2022-05-19 09:37:23 Trying 172.16.0.20:22
2022-05-19 09:37:27 STOP
2022-05-19 09:37:27 No results found...
2022-05-19 09:37:27 START
2022-05-19 09:37:27 Crowbar v0.4.3-dev
2022-05-19 09:37:27 Trying 172.16.0.20:22
2022-05-19 09:37:31 STOP
2022-05-19 09:37:31 No results found...
2022-05-19 09:37:31 START
2022-05-19 09:37:31 Crowbar v0.4.3-dev
2022-05-19 09:37:31 Trying 172.16.0.20:22
2022-05-19 09:37:34 STOP
2022-05-19 09:37:34 No results found...
2022-05-19 09:37:35 START
2022-05-19 09:37:35 Crowbar v0.4.3-dev
2022-05-19 09:37:35 Trying 172.16.0.20:22
2022-05-19 09:37:38 STOP
2022-05-19 09:37:38 No results found...
2022-05-19 09:37:38 START
2022-05-19 09:37:38 Crowbar v0.4.3-dev
2022-05-19 09:37:38 Trying 172.16.0.20:22
2022-05-19 09:37:42 STOP
2022-05-19 09:37:42 No results found...
2022-05-19 09:37:42 START
2022-05-19 09:37:42 Crowbar v0.4.3-dev
2022-05-19 09:37:42 Trying 172.16.0.20:22
2022-05-19 09:37:46 STOP
2022-05-19 09:37:46 No results found...
-----------------------------------------------------------------------
***********************************************************************
* Test #3 - RDP Brute Force with Password List *
* This simulates an RDP brute force attack on the internal RDP port *
* of the windows server that we installed in the environment. It uses*
* a list of common passwords that can be found on the web. This test *
* will trigger a detection, but will fail to get into the target *
* windows instance. *
***********************************************************************
Sending 250 password attempts at the windows server...
Hydra v9.4-dev (c) 2022 by van Hauser/THC & David Maciejak - Please do not use in military or secret service organizations, or for illegal purposes (this is non-binding, these *** ignore laws and ethics anyway).
Hydra (https://github.com/vanhauser-thc/thc-hydra) starting at 2022-05-19 09:37:46
[WARNING] rdp servers often don't like many connections, use -t 1 or -t 4 to reduce the number of parallel connections and -W 1 or -W 3 to wait between connection to allow the server to recover
[INFO] Reduced number of tasks to 4 (rdp does not like many parallel connections)
[WARNING] the rdp module is experimental. Please test, report - and if possible, fix.
[DATA] max 4 tasks per 1 server, overall 4 tasks, 1792 login tries (l:7/p:256), ~448 tries per task
[DATA] attacking rdp://172.16.0.24:3389/
[STATUS] 1099.00 tries/min, 1099 tries in 00:01h, 693 to do in 00:01h, 4 active
1 of 1 target completed, 0 valid password found
Hydra (https://github.com/vanhauser-thc/thc-hydra) finished at 2022-05-19 09:39:23
-----------------------------------------------------------------------
***********************************************************************
* Test #4 - CryptoCurrency Mining Activity *
* This simulates interaction with a cryptocurrency mining pool which *
* can be an indication of an instance compromise. In this case, we are*
* only interacting with the URL of the pool, but not downloading *
* any files. This will trigger a threat intel based detection. *
***********************************************************************
Calling bitcoin wallets to download mining toolkits
-----------------------------------------------------------------------
***********************************************************************
* Test #5 - DNS Exfiltration *
* A common exfiltration technique is to tunnel data out over DNS *
* to a fake domain. Its an effective technique because most hosts *
* have outbound DNS ports open. This test wont exfiltrate any data, *
* but it will generate enough unusual DNS activity to trigger the *
* detection. *
***********************************************************************
Calling large numbers of large domains to simulate tunneling via DNS
***********************************************************************
* Test #6 - Fake domain to prove that GuardDuty is working *
* This is a permanent fake domain that customers can use to prove that*
* GuardDuty is working. Calling this domain will always generate the *
* Backdoor:EC2/C&CActivity.B!DNS finding type *
***********************************************************************
Calling a well known fake domain that is used to generate a known finding
; <<>> DiG 9.11.4-P2-RedHat-9.11.4-26.P2.amzn2.5.2 <<>> GuardDutyC2ActivityB.com any
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 11495
;; flags: qr rd ra; QUERY: 1, ANSWER: 8, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
;; QUESTION SECTION:
;GuardDutyC2ActivityB.com. IN ANY
;; ANSWER SECTION:
GuardDutyC2ActivityB.com. 6943 IN SOA ns1.markmonitor.com. hostmaster.markmonitor.com. 2018091906 86400 3600 2592000 172800
GuardDutyC2ActivityB.com. 6943 IN NS ns3.markmonitor.com.
GuardDutyC2ActivityB.com. 6943 IN NS ns5.markmonitor.com.
GuardDutyC2ActivityB.com. 6943 IN NS ns7.markmonitor.com.
GuardDutyC2ActivityB.com. 6943 IN NS ns2.markmonitor.com.
GuardDutyC2ActivityB.com. 6943 IN NS ns4.markmonitor.com.
GuardDutyC2ActivityB.com. 6943 IN NS ns6.markmonitor.com.
GuardDutyC2ActivityB.com. 6943 IN NS ns1.markmonitor.com.
;; Query time: 27 msec
;; SERVER: 172.16.0.2#53(172.16.0.2)
;; WHEN: Thu May 19 09:39:23 UTC 2022
;; MSG SIZE rcvd: 238
*****************************************************************************************************
Expected GuardDuty Findings
Test 1: Internal Port Scanning
Expected Finding: EC2 Instance i-011e73af27562827b is performing outbound port scans against remote host. 172.16.0.20
Finding Type: Recon:EC2/Portscan
Test 2: SSH Brute Force with Compromised Keys
Expecting two findings - one for the outbound and one for the inbound detection
Outbound: i-011e73af27562827b is performing SSH brute force attacks against 172.16.0.20
Inbound: 172.16.0.25 is performing SSH brute force attacks against i-0bada13e0aa12d383
Finding Type: UnauthorizedAccess:EC2/SSHBruteForce
Test 3: RDP Brute Force with Password List
Expecting two findings - one for the outbound and one for the inbound detection
Outbound: i-011e73af27562827b is performing RDP brute force attacks against 172.16.0.24
Inbound: 172.16.0.25 is performing RDP brute force attacks against i-0191573dec3b66924
Finding Type : UnauthorizedAccess:EC2/RDPBruteForce
Test 4: Cryptocurrency Activity
Expected Finding: EC2 Instance i-011e73af27562827b is querying a domain name that is associated with bitcoin activity
Finding Type : CryptoCurrency:EC2/BitcoinTool.B!DNS
Test 5: DNS Exfiltration
Expected Finding: EC2 instance i-011e73af27562827b is attempting to query domain names that resemble exfiltrated data
Finding Type : Trojan:EC2/DNSDataExfiltration
Test 6: C&C Activity
Expected Finding: EC2 instance i-011e73af27562827b is querying a domain name associated with a known Command & Control server.
Finding Type : Backdoor:EC2/C&CActivity.B!DNS
After a few minutes, the findings appear in the GuardDuty console. At the top, I see the malicious files found by the new Malware Protection capability. One of the findings is related to an EC2 instance, the other to an ECS cluster.
First, I select the finding related to the EC2 instance. In the panel, I see the information on the instance and the malicious file, such as the file name and path. In the Malware scan details section, the Trigger finding ID points to the original GuardDuty finding that triggered the malware scan. In my case, the original finding was that this EC2 instance was performing RDP brute force attacks against another EC2 instance.
Here, I choose Investigate with Detective and, directly from the GuardDuty console, I go to the Detective console to visualize AWS CloudTrail and Amazon Virtual Private Cloud (Amazon VPC) flow data for the EC2 instance, the AWS account, and the IP address affected by the finding. Using Detective, I can analyze, investigate, and identify the root cause of suspicious activities found by GuardDuty.
When I select the finding related to the ECS cluster, I have more information on the resource affected, such as the details of the ECS cluster, the task, the containers, and the container images.
Using the GuardDuty tester scripts makes it easier to test the overall integration of GuardDuty with other security frameworks you use so that you can be ready when a real threat is detected.
Comparing GuardDuty Malware Protection with Amazon Inspector
At this point, you might ask yourself how GuardDuty Malware Protection relates to Amazon Inspector, a service that scans AWS workloads for software vulnerabilities and unintended network exposure. The two services complement each other and offer different layers of protection:
- Amazon Inspector offers proactive protection by identifying known software and application vulnerabilities that serve as an entry point for attackers to compromise resources and install malware.
- GuardDuty Malware Protection detects malware that is found to be present on actively running workloads. At that point, the system has already been compromised, but GuardDuty can limit the time of an infection and take action before a system compromise results in a business-impacting event.
Availability and Pricing
Amazon GuardDuty Malware Protection is available today in all AWS Regions where GuardDuty is available, excluding the AWS China (Beijing), AWS China (Ningxia), AWS GovCloud (US-East), and AWS GovCloud (US-West) Regions.
At launch, GuardDuty Malware Protection is integrated with these partner offerings:
With GuardDuty, you don’t need to deploy security software or agents to monitor for malware. You only pay for the amount of GB scanned in the file systems (not for the size of the EBS volumes) and for the EBS snapshots during the time they are kept in your account. All EBS snapshots created by GuardDuty are automatically deleted after they are scanned unless you enable snapshot retention when malware is found. For more information, see GuardDuty pricing and EBS pricing. Note that GuardDuty only scans EBS volumes less than 1 TB in size. To help you control costs and avoid repeating alarms, the same volume is not scanned more often than once every 24 hours.
Detect malicious activity and protect your applications from malware with Amazon GuardDuty.
— Danilo
