Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
An OAuth consent hands an application standing access to mail, files, and chat without a password. Auditing consents means listing every application, reading its permissions, and removing what nobody can justify.

How Do You Audit OAuth App Consents?
When a user clicks “Accept” on an application’s permission request, they issue an OAuth consent: a standing grant that lets that application call Microsoft Graph or the Google APIs on their behalf, using refresh tokens rather than a password. Auditing consents means producing a full inventory of which applications hold which permissions over whose data, judging each one, and revoking the rest. It matters because a consent is unaffected by a password reset and, in most configurations, unaffected by multi-factor authentication once granted.
An illicit consent grant attack skips the hard parts of account takeover. The victim is sent to a legitimate Microsoft or Google sign-in page — the URL and the certificate are genuine — and is asked to approve an application with an unremarkable name. There is no credential to phish, no second factor to intercept, and no anomalous sign-in to detect, because the attacker never signs in as the user at all.
The pattern has a long history. On 3 May 2017 a worm posing as Google Docs spread through roughly a million Google accounts in an hour by requesting mailbox and contacts scopes. In April 2022, stolen OAuth tokens issued to the Heroku and Travis CI integrations were used to clone private repositories from GitHub customers. And in the intrusion Microsoft disclosed in January 2024, the attackers pivoted from a legacy non-production tenant by granting a malicious application the full_access_as_app role over Exchange Online, then read corporate mailboxes with it. In every case the access was an application permission, not a stolen password.
The distinction decides how dangerous a grant is.
| Delegated | Application | |
|---|---|---|
| Acts as | The signed-in user | Itself, with no user |
| Scope limit | What that user can reach | Everything the permission covers, tenant-wide |
| Consent | User or administrator | Administrator only |
| Example | Mail.Read for one mailbox | Mail.ReadWrite for every mailbox |
Application permissions such as Mail.ReadWrite, Files.ReadWrite.All, Directory.ReadWrite.All, and RoleManagement.ReadWrite.Directory deserve individual justification and an owner. Treat offline_access as significant too: it is what turns a one-off approval into persistent access through refresh tokens.
Auditing is remedial. The durable fix is to stop unrestricted user consent. In Entra ID, set user consent to allow only verified publishers requesting low-impact permissions, and enable the admin consent request workflow so legitimate requests reach an administrator rather than being refused silently. Google Workspace supports the same shape of control through app access restrictions and an allow list of trusted applications. Both platforms will flag suspicious applications automatically when the relevant detection features are licensed, and those alerts should route to your security operations centre rather than to an unread mailbox.
Build the review into a cycle: monthly for new grants, quarterly for the full inventory, and immediately after any phishing campaign that reached inboxes. This sits naturally alongside wider SaaS security posture management and your Microsoft 365 configuration baseline, and it is one of the few controls that directly addresses shadow IT, since every self-service SaaS tool a user adopts leaves a consent record behind.
Under UK GDPR, an application reading mailboxes is processing personal data, usually without a contract under Article 28. That makes the consent inventory a data protection artefact as much as a security one.
Not quite. Deleting the service principal or the grant prevents new tokens from being issued, but an access token already in the attacker’s hands stays valid until it expires, typically within an hour. For a confirmed compromise, revoke refresh tokens and sign-in sessions for the affected users as well, then watch service principal sign-in logs to confirm the activity has stopped.
Blocking it outright is the safest setting but creates friction that pushes people towards unmanaged alternatives. The middle path works better: allow consent only for applications from verified publishers requesting low-impact permissions, and enable the admin consent request workflow so users can ask for anything else. Respond to those requests quickly, or the policy will be worked around.
Look for an application registered inside your own tenant by a standard user, with no listed owner, no verified publisher, a generic or lookalike name, a reply URL pointing at an unfamiliar domain, and permissions far wider than its stated purpose. A creation timestamp that lines up with a phishing wave is strong corroboration. Never-used grants and heavily used unknown grants both warrant investigation.
Effectively, yes, after the initial approval. The user completes authentication once, including any second factor, and the application then uses refresh tokens for ongoing access without further prompts. Conditional access policies can be scoped to applications, and continuous access evaluation shortens the window after a revocation, but a granted consent should be treated as access that survives your authentication controls.
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.