Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

How Do You Collect Evidence in SaaS Environments?

In a SaaS tenant there is nothing to image, so the provider’s audit log is the entire evidence base, and what it contains depends on licensing decisions made long before the incident.

Illustration for the cybersecurity term SaaS Evidence Collection
Home / Blog / Cybersecurity Terms / How Do You Collect Evidence in SaaS Environments?
17 November 2026

How Do You Collect Evidence in SaaS Environments?

How Do You Collect Evidence in SaaS Environments?

When an attacker compromises a mailbox or a file sharing tenant, there is no disk to image and no memory to capture. The only record of what happened is the log the provider chose to keep, for the period their retention policy allows, at the level of detail your licence entitles you to. SaaS evidence collection is therefore less about technique and more about knowing in advance what will exist.

The uncomfortable version of that statement: the decision about whether you can investigate a mailbox compromise was made by whoever signed the licensing agreement.

Preserve Before You Investigate

The first action is always preservation, because provider retention windows run from the event and not from your discovery of it. In Microsoft 365, apply a Purview eDiscovery hold to the affected mailboxes and sites. In Google Workspace, create a Vault matter and hold. In both cases this stops deletion, including deletion by an attacker who still has access, while you work out the scope.

Only then begin collection. Preservation is cheap and reversible; a log that has aged out is gone.

Microsoft 365

The Unified Audit Log is the primary source. Search it through Purview or, for anything substantial, through PowerShell:

Search-UnifiedAuditLog -StartDate 01/09/2026 -EndDate 15/09/2026 `
  -UserIds [email protected] -ResultSize 5000

Retention is 180 days for most audit records under Audit (Standard) and one year under Audit (Premium), with longer retention available as an add-on. The record that decides most mailbox compromise cases is MailItemsAccessed, which shows which messages were actually read rather than merely that a session existed. Microsoft extended availability of this record to lower tiers after the Storm-0558 intrusion in 2023, but confirm it is enabled in your tenant rather than assuming.

Collect alongside it: Entra ID sign-in logs, which default to short retention unless exported to Log Analytics or a SIEM; inbox rules and forwarding configuration, a persistent favourite for business email compromise; OAuth application consents and service principal credentials; and message trace data for what was sent.

Other Platforms

Making the Export Defensible

  1. Export through documented API queries or the provider’s own export function, and record the exact query, the time window, and the time zone the service used.
  2. Hash every exported file on receipt and store the values separately.
  3. Normalise all timestamps to UTC before building a timeline, since providers differ in what they return.
  4. Record the licence tier and audit configuration in force during the incident window, because it defines what the absence of a record means.
  5. Note that absence of evidence is not evidence of absence. If MailItemsAccessed was not available, you cannot say the attacker did not read the mailbox, and the ICO notification should say so honestly.

The Shared Responsibility Angle

Under UK GDPR the provider is your processor and you remain the controller, so the duty to investigate and notify is yours even though the evidence sits in their infrastructure. Article 28 requires the contract to oblige the processor to assist you with breach notification and to make information available. That is a clause worth reading before signing: what logs are available, for how long, in what format, and how quickly the provider will respond to a forensic request.

Practical preparation beats contractual rights. Stream identity and audit logs into your own SIEM continuously, so the evidence exists on your side of the boundary with a retention period you control, and pair it with SaaS security posture management to spot the configuration drift that creates the exposure in the first place.

At Privia Security, SaaS log collection forms part of the first hour of any Investigation and Incident Response engagement involving cloud identity.

FAQ – Frequently Asked Questions

How Long Does Microsoft 365 Keep Audit Logs?

Most audit records are retained for 180 days under Audit (Standard) and one year under Audit (Premium), with a ten-year add-on available for specific record types. Entra ID sign-in log retention is much shorter by default. Because these windows run from the event rather than from discovery, apply holds and export the data as soon as an incident is suspected.

What Is MailItemsAccessed and Why Does It Matter?

It is a Microsoft 365 audit record showing which individual messages were accessed in a mailbox, rather than merely that a sign-in occurred. In a mailbox compromise it is the difference between telling the ICO which personal data was exposed and admitting you cannot tell. Confirm it is enabled in your tenant before you need it, since availability depends on configuration and licensing.

Can You Image a SaaS Tenant Like a Server?

No. You have no access to the underlying storage or memory, and the provider will not supply it. The equivalent of an image is a complete, hash-verified export of the relevant audit logs, mailbox or file content, and configuration state, collected through documented API queries. This is why preservation holds and continuous log export matter more than any acquisition tool.

Who Is Responsible for Investigating a SaaS Breach?

You are. Under UK GDPR the SaaS provider is normally your processor and your organisation remains the controller, so the duty to assess the breach and notify the ICO within 72 hours sits with you. The provider must assist, and Article 28 requires the contract to say so, but the regulatory obligation and the reputational consequences are not transferable.

Related Terms

You May Be Interested In These

What Is SOAR?

Cybersecurity Terms

17 November 2026

What Is SOAR?

Read More
What Is an Intrusion Detection System (IDS)?

Cybersecurity Terms

17 November 2026

What Is an Intrusion Detection System (IDS)?

Read More
What Is a Supply Chain Attack?

Cybersecurity Terms

17 November 2026

What Is a Supply Chain Attack?

Read More