Cyber forensics
Influence the future of the AWS Security Reference Architecture (AWS SRA) by taking a short survey |
In the context of the AWS SRA, we use the following definition of forensics provided by
the National Institute of Standards and Technology (NIST): "the application of science to the
identification, collection, examination, and analysis of data while preserving the integrity
of the information and maintaining a strict chain of custody for the data" (source: NIST
Special Publication 800-86 – Guide to Integrating Forensic Techniques into Incident
Response
Forensics in the context of security incident response
The incident response (IR) guidance in this section is provided only in the context of forensics and how different services and solutions can improve the IR process.
The AWS Security Incident Response Guide lists best practices for responding to
security incidents in the AWS Cloud, based on the experiences of the AWS Customer Incident Response Team (AWS CIRT)
The National Institute of
Standards and Technology Cybersecurity Framework (NIST CSF)
This repeated cycle of analysis, containment, eradication, and back to analysis again allows you to gather more information each time new indicators of compromise (IoCs) are detected. Those IoCs are useful from a number of perspectives. They provide you with a story of the steps that were taken by the adversary in order to compromise your environment. Also, by performing proper post-incident review, you can improve your defenses and detections so that you can prevent the incident in the future or detect the adversary’s actions faster and thus reduce the impact of the incident.
Although this IR process isn’t the main objective of forensics, many of the tools, techniques, and best practices are shared with IR (especially the analysis step). For example, after the detection of an incident, the forensic collection process gathers the evidence. Next, evidence examination and analysis can help to extract IoCs. At the end, forensic reporting can assist in post-IR activities.
We recommend that you automate the forensic process as much as possible to speed up the response and reduce the load on IR stakeholders. In addition, you can add more automated analyses after the forensic collection process has finished and the evidence has been securely stored to avoid contamination. For more information, see the pattern Automate incident response and forensics on the AWS Prescriptive Guidance website.
Design considerations
To improve your security IR preparedness:
-
Enable and securely store logs that might be required during an investigation or incident response.
-
Prebuild queries for known scenarios and provide automated ways to search logs. Consider using Amazon Detective.
-
Prepare your IR tooling by running simulations.
-
Regularly test backup and recovery processes to make sure they are successful.
-
Use scenario-based playbooks, starting with common potential events related to AWS based on Amazon GuardDuty findings. For information about how to build your own playbooks, see the Playbook resources section of the AWS Security Incident Response Guide.
Forensics account
Disclaimer
The following description of an AWS Forensics account should only be used by organizations as a starting point for organizations to develop their own forensic capabilities in conjunction with guidance from their legal advisors.
We make no claim as to the suitability of this guidance in the detection or investigation of crime, nor the ability of data or forensics evidence captured through the application of this guidance to be used in a court of law. You should independently evaluate the suitability of best practices described here for your use case.
The following diagram illustrates the AWS security services that can be configured in a dedicated Forensics account. For context, the diagram shows the Security Tooling account to depict the AWS services that are used to provide detection or notifications in the Forensics account.
![Forensics account on AWS](/images/prescriptive-guidance/latest/security-reference-architecture/images/forensics-acct.png)
The Forensics account is a separate and dedicated type of Security Tooling account that is within the Security OU. The purpose of the Forensics account is to provide a standard, pre-configured, and repeatable clean room to allow an organization’s forensics team to implement all phases of the forensics process: collection, examination, analysis, and reporting. In addition, the quarantine and isolation process for in-scope resources are also included in this account.
Containing the entire forensics process in a separate account allows you to apply additional access controls to the forensic data that’s collected and stored. We recommend that you separate the Forensics and Security Tooling accounts for the following reasons:
-
Forensics and security resources might be on different teams or have different permissions.
-
The Security Tooling account might have automation that’s focused on responding to security events at the AWS control plane, such as enabling Amazon S3 Block Public Access for S3 buckets, whereas the Forensics account also includes AWS data plane artifacts that the customer might be responsible for, such as operating system (OS) or application-specific data within an EC2 instance.
-
You might need to implement additional access restrictions or legal holds depending on your organizational or regulatory requirements.
-
The forensic analysis process might require analysis of malicious code such as malware in a secured environment in alignment with the AWS terms of service.
The Forensics account should include automation to expedite evidence collection at scale while minimizing human interaction in the forensic collection process. The automation to respond and quarantine resources would also be included in this account to simplify tracking and reporting mechanisms.
The forensic capabilities described in this section should be deployed into every available AWS Region, even if your organization isn’t actively using the capabilities. If you don’t plan to use specific AWS Regions, you should apply a service control policy (SCP) to restrict provisioning AWS resources. Additionally, maintaining investigations and storage of forensic artifacts within the same Region helps avoid issues with the changing regulatory landscape of data residency and ownership.
This guidance uses the Log Archive account as outlined previously to record actions taken in the environment through AWS APIs, including the APIs that you run in the Forensics account. Having such logs can help avoid allegations of mishandling or tampering of artifacts. Depending on the level of detail that you enable (see Logging management events and Logging data events in the AWS CloudTrail documentation), the logs can include information about the account used to collect the artifacts, the time the artifacts were collected, and the steps taken to collect the data. By storing artifacts in Amazon S3, you can also use advanced access controls and log information about who had access to the objects. A detailed log of actions allows others to repeat the process later if needed (assuming that the resources in scope are still available).
Design considerations
-
Automation is helpful when you have many concurrent incidents, because it helps speed up and scale the collection of vital evidence. However, you should consider these benefits carefully. For example, in case of a false positive incident, a fully automated forensic response might negatively impact a business process that’s supported by an AWS workload in scope. For more information, see the design considerations for AWS GuardDuty, AWS Security Hub, and AWS Step Functions in the following sections.
-
We recommend separate Security Tooling and Forensics accounts, even if your organization’s forensics and security resources are on the same team and all functions can be performed by any member of the team. Splitting the functions into separate accounts further supports least privilege, helps avoid contamination from an ongoing security event analysis, and helps enforce the integrity of artifacts that are gathered.
-
You can create a separate Forensics OU to host this account if you want to further emphasize the separation of duties, least privilege, and restrictive guardrails.
-
If your organization uses immutable infrastructure resources, information that is forensically valuable might get lost if a resource is automatically deleted (for example, during a scaling down event) and before a security incident is detected. To avoid this, consider running a forensic collection process for each such resource. To reduce the volume of data collected, you can consider factors such as environments, business criticality of the workload, type of data processed, and so on.
-
Consider using Amazon WorkSpaces to spin up clean workstations. This can help separate actions of stakeholders during an investigation.
Amazon GuardDuty
Amazon GuardDuty
You can use GuardDuty findings to initiate the forensic workflow that captures disk and memory images of potentially compromised EC2 instances. This reduces human interaction and can significantly increase the speed of forensic data collection. You can integrate GuardDuty with Amazon EventBridge to automate responses to new GuardDuty findings.
The list of GuardDuty finding types is growing. You should consider which finding types (for example, Amazon EC2, Amazon EKS, malware protection, and so on) should initiate the forensic workflow.
You can fully automate the integration of the containment and forensic data collection process with GuardDuty findings to capture the investigation of disk and memory artifacts and quarantine EC2 instances. For example, if all ingress and egress rules are removed from a security group, you can apply a network ACL to interrupt the existing connection and attach an IAM policy to deny all requests.
Design considerations
-
Depending on the AWS service, the customer’s shared responsibility can vary. For example, capturing volatile data on EC2 instances is possible only on the instance itself, and might include valuable data that can be used as forensic evidence. Conversely, responding and investigating a finding for Amazon S3 primarily involves CloudTrail data or Amazon S3 access logs. Response automation should be organized across both the Security Tooling and Forensics accounts depending on the customer’s shared responsibility, the general process flow, and the captured artifacts that need to be secured.
-
Before you quarantine an EC2 instance, weigh its overall business impact and criticality. Consider establishing a process where appropriate stakeholders are consulted before you use automation to contain the EC2 instance.
AWS Security Hub
AWS Security Hub
In addition to monitoring your security posture, Security Hub supports integration with Amazon EventBridge to automate the remediation of specific findings. For example, you can define custom actions that can be programmed to run an AWS Lambda function or an AWS Step Functions workflow to implement a forensic process.
Security Hub custom actions provide a standardized mechanism for authorized security analysts or resources to implement containment and forensic automation. This reduces human interactions in the containment and capture of forensic evidence. You can add a manual checkpoint in the automated process to confirm that a forensic collection is actually required.
Design consideration
-
Security Hub can be integrated with many services, including AWS Partner solutions. If your organization uses detective security controls that aren’t fully fine-tuned and sometimes result in false positive alerts, fully automating the forensic collection process would result in running that process unnecessarily.
Amazon EventBridge
Amazon EventBridge
For example, you can use EventBridge as a mechanism to initiate a forensic workflow in Step Functions to capture disk and memory images based on detections from security monitoring tools such as GuardDuty. Or you could use it in a more manual way: EventBridge could detect tag change events in CloudTrail, which could initiate the forensic workflow in Step Functions.
AWS Step Functions
AWS Step Functions
Step Functions is ideal for use with a forensic process because it supports a repeatable, automated set of predefined steps that can be verified through AWS logs. This helps you exclude any human involvement and avoid mistakes in your forensic process.
Design considerations
-
You can initiate a Step Functions workflow manually or automatically to capture and analyze security data when GuardDuty or Security Hub indicates a compromise. Automation with minimal or no human interaction enables your team to quickly scale in case of a significant security event that affects many resources.
-
To limit fully automated workflows, you can include steps in the automation flow for some manual intervention. For example, you might require an authorized security analyst or team member to review the generated security findings and determine whether to initiate a collection of forensic evidence, or quarantine and contain affected resources, or both.
-
If you want to initiate a forensic investigation without an active finding created from security tooling (such as GuardDuty or Security Hub), you should implement additional integrations to invoke a forensic Step Functions workflow. This can be done by creating an EventBridge rule that looks for a specific CloudTrail event (such as a tag change event) or by allowing a security analyst or team member to start a forensic Step Functions workflow directly from the console. You can also use Step Functions to create actionable tickets by integrating it with your organization’s ticketing system.
AWS Lambda
With AWS Lambda
In the context of a forensic investigation, using Lambda functions helps you achieve constant results through repeatable, automated, and predefined steps that are defined in the Lambda code. When a Lambda function runs, it creates a log that helps you verify that the proper process was implemented.
Design considerations
-
Lambda functions have a timeout of 15 minutes, whereas a comprehensive forensic process to collect relevant evidence might take longer. For this reason, we recommend that you orchestrate your forensic process by using Lambda functions that are integrated in a Step Functions workflow. The workflow lets you create Lambda functions in the correct order, and each Lambda function implements an individual collection step.
-
By organizing your forensic Lambda functions into a Step Functions workflow, you can run parts of the forensic collection procedure in parallel to speed up the collection. For example, you can collect information about the creation of disk images faster when multiple volumes are in scope.
AWS KMS
AWS Key Management Service
As part of the forensics process, data collection and investigation should be done in an isolated environment to minimize business impact. Data security and integrity cannot be compromised during this process, and a process will need to be put in place to allow sharing of encrypted resources, such as snapshots and disk volumes, between the potentially compromised account and the Forensics account. In order to accomplish this, your organization will have to make sure that the associated AWS KMS resource policy supports reading the encrypted data as well as securing the data by re-encrypting it with an AWS KMS key in the Forensics account.
Design consideration
-
An organization’s KMS key policies should allow authorized IAM principals for forensics to use the key to decrypt data in the source account and re-encrypt it in the Forensics account. Use infrastructure as code (IaC) to centrally manage all your organization’s keys in AWS KMS to help ensure that only authorized IAM principals have the appropriate and least privilege access. These permissions should exist on all KMS keys that can be used to encrypt resources on AWS that could be collected during a forensics investigation. If you update the KMS key policy after a security event, the subsequent resource policy update for a KMS key that’s in use might impact your business. Additionally, permission issues can increase the overall mean time to respond (MTTR) for a security event.