Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Cloud forensics investigates incidents in environments you do not physically control, where evidence is an API call away, retention is a billing decision, and the hardware belongs to someone else.

What Is Cloud Forensics?
Cloud forensics is digital forensics applied to infrastructure and services you rent rather than own. The analytical questions are the same ones you would ask about a server room — what happened, when, by whom, and what did they take — but almost every assumption underneath has changed. You cannot seize the hardware, the instance you want to image may have been terminated by an autoscaler an hour ago, and the most valuable evidence is often a management API log rather than a disk.
In a traditional intrusion the attacker works inside the operating system. In the cloud they more often work against the management layer, using stolen access keys or a compromised identity to enumerate resources, create new credentials, disable logging, snapshot a database, and share that snapshot with an account they control. None of that touches a guest operating system, so a disk image tells you nothing.
What tells you everything is the control plane audit trail: CloudTrail in AWS, the Activity Log and Microsoft Entra sign-in and audit logs in Azure, Cloud Audit Logs in Google Cloud. Learn their defaults before you need them. CloudTrail management events are retained in console event history for 90 days unless a trail writes them to storage. Entra sign-in logs are kept for 30 days without a premium licence. Microsoft 365 audit records sit behind a licensing tier that determines whether you get 180 days or a year. These are the single most common reason an investigation stops short of an answer.
The usual sequence for a compromised virtual machine is to isolate it with a restrictive security group rather than stopping it, capture memory through the cloud agent or a run command, take a snapshot of each attached volume, copy the snapshot into a dedicated forensic account, and attach it read-only to an analysis instance. Snapshots are themselves evidence: record the identifier, the creation time, and a hash of the exported image, and keep the chain of custody as carefully as you would for a physical disk.
Beyond the virtual machine, there are sources with no on-premises equivalent: object storage access logs, serverless function invocation records, managed database audit logs, VPC flow logs, DNS query logs, and the instance metadata service, which is itself a frequent target — see IMDS attacks for why. Containers add another layer, covered in Kubernetes incident analysis.
The provider secures and logs the platform; you secure and log what you run on it. That boundary also defines what you can obtain. A hypervisor log, the physical host, or another tenant’s activity is not available to you, and a provider will generally respond to a lawful request from a law enforcement agency rather than from a customer’s investigator. Plan your evidence strategy around what your own account can produce, because that is effectively all of it.
Data held in one region may be replicated to another, and the legal route to obtain it differs accordingly. The UK–US Data Access Agreement, in force since 3 October 2022, allows UK authorities to serve requests directly on US communications providers in serious crime cases, which is considerably faster than a mutual legal assistance treaty request. For a corporate investigation, the more pressing question is usually UK GDPR: an export of a mailbox or a storage bucket often sweeps in personal data belonging to people who are not under suspicion, so scope the collection and document the lawful basis.
NIST Internal Report 8006 remains the clearest catalogue of why cloud investigations fail, and most of its findings reduce to preparation rather than technique.
Our cloud security services include reviewing whether the evidence you would need actually exists in your tenant today, which is a cheaper discovery than making it during an incident.
No. You never obtain physical custody of hardware in a shared environment, and a provider will not hand over a host that serves multiple tenants. The equivalent is a volume snapshot plus memory capture taken through your own account, supported by control plane logs. Law enforcement can compel disclosure through legal process, but a customer cannot.
Less time than most teams assume. AWS CloudTrail event history covers 90 days in the console unless a trail stores events elsewhere, Microsoft Entra sign-in logs are kept 30 days without premium licensing, and Microsoft 365 audit retention depends on the subscription tier. Check these settings now and extend them, because retention cannot be applied retrospectively.
Isolate it first. Stopping an instance discards memory, which may hold the only copy of injected code, decryption keys, and active network connections. Replacing its security group with a deny-all rule removes the attacker’s access while preserving volatile evidence, after which you can capture memory and snapshot the volumes in a controlled sequence.
Up to a point. You get invocation records, application logs, and the function’s own output, but the execution environment is gone within minutes and there is no disk to image. That makes detailed logging inside the function and strict least-privilege roles far more important than in a virtual machine estate, because the telemetry you did not emit cannot be recovered afterwards.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.