Microsoft 365

安全發現結果

每次 Microsoft 365 掃描時會執行 26 個偵測規則。每個發現結果都附帶純文字說明、背後的證據和修復步驟。

• CRITICALPublic exposure

  Public link with edit/owner access

• CRITICALExternal access

  External principal holds ownership

• CRITICALIdentity

  A material share of users have not registered for multi-factor authentication

  Users who have never registered an MFA method cannot be challenged for a second factor, so any policy requiring MFA silently fails for them. Aim for near-complete registration.

• HIGHPublic exposure

  Public link share (link-only)

• HIGHPublic exposure

  Public + web-discoverable share

• HIGHExternal access

  External access to sensitive data

• HIGHExternal access

  External collaborator with edit access

• HIGHOver permission

  Sensitive resource editable org-wide

• HIGHOver permission

  Broad group access to sensitive data

• HIGHAI agent

  Service account with broad write access

• HIGHAI agent

  AI agent with access to sensitive data

• HIGHAI agent

  AI agent with edit/owner access

• HIGHExternal access

  Access granted to a look-alike (typosquat) domain

• HIGHIdentity

  Security Defaults are off and no Conditional Access baseline is enforced

  A tenant with Security Defaults disabled and no enabled Conditional Access policy has no baseline MFA/identity protection — a single phished password can sign in unchallenged.

• HIGHAccess

  Legacy authentication is not blocked

  Legacy authentication protocols (IMAP, POP, SMTP AUTH, older Office clients) cannot enforce MFA and are the vector for the majority of password-spray compromises. An enabled Conditional Access policy should block them.

• HIGHAccess

  No enabled Conditional Access policy requires multi-factor authentication

  Without an enforced policy requiring MFA, users authenticate with a password alone. Report-only and disabled policies do not enforce anything.

• HIGHApplications

  Users can consent to third-party apps themselves

  When user consent is left enabled, any employee can grant a third-party app (including an AI assistant) access to their mailbox or files with one click — the primary shadow-AI / OAuth-phishing vector. Route consent through an admin review workflow instead.

• MEDIUMOver permission

  Broad org-wide edit access (aggregate)

• MEDIUMOver permission

  Resource with multiple owners

• MEDIUMStale access

  Stale external read access

• MEDIUMStale access

  Dormant privileged internal account

• MEDIUMMisconfig

  Direct share breaks folder inheritance

• MEDIUMAccess

  Anyone in the organization can invite external guests

  When every member (not just admins or designated inviters) can invite B2B guests, external identities accumulate without review — a quiet expansion of the access boundary.

• MEDIUMSharing

  SharePoint / OneDrive allows anonymous "Anyone" sharing links org-wide

  When the tenant external-sharing capability permits "Anyone" links, any user can mint a no-sign-in URL to a document, and that URL leaks data the moment it is forwarded.

• MEDIUMAdmin

  More Global Administrators than recommended

  Global Administrator is the most powerful role in the tenant. Microsoft recommends keeping it to a small number (around 2–4, with break-glass accounts excluded). Every extra holder is a high-value phishing target with tenant-wide blast radius.

• MEDIUMMisconfig

  Vendor AI-training data-contribution posture