SaaS Security

Video guide: Find SaaS security setup and coverage checks

Recorded October 22, 2026 in the fictional Summit Managed IT staging workspace, App 0.432.1. The deployed demo can differ from the current product; use the written guide for current behavior. No customer has a SaaS connection in the demo; no incident view, collector run, response approval or provider recovery is demonstrated.

Video transcript

Open SaaS Security. No customer has a SaaS connection in this fictional workspace, so the page shows a setup prerequisite. Counts without current feeds would be incomplete evidence. This walkthrough does not establish that any customer connection is healthy.

Open the written setup guide because no customer incident view is available in the unconnected demo. Supported collection and response actions differ by provider. Missing feeds, stale samples and disconnected providers must remain explicit coverage gaps; an empty result is not proof of no compromise.

Use the written guide for the implemented providers and Microsoft-specific response limits. Verify a successful collector run, customer mapping and an authorized benign signal. Review the exact response plan, then check every provider result and recovery step independently.

What it does

SaaS Security is continuous detection and controlled response for the customer's cloud tenants: it collects the tenant's events (sign-ins, directory audits, Exchange, SharePoint and Azure AD unified audit log for Microsoft 365, and provider-specific feeds for the 18 wired generic integrations listed below), normalizes them into one event store, evaluates versioned detection rules, opens incidents with an audited lifecycle, supports investigations and simulations, maps cloud accounts to managed devices, keeps posture and hardening findings, and can execute response plans (revoke sessions) and configuration change plans (security defaults, SharePoint sharing, user consent) under an enforced per-customer response policy with a kill switch.

How it works

  • Capabilities: on connect, every data feed is probed and marked available, permission missing, license missing, unsupported, unverified or degraded, so the Integrations tab says exactly what each tenant can and cannot deliver (for example sign-in logs are license missing on tenants without Entra ID P1; logons then come from the unified audit log).
  • Capability checks refresh during collection when their result is older than six hours. Integrations marks older results stale; an administrator can re-check capabilities at any time.
  • Collection: the saas-collect timer runs every 5 minutes per connection with a claim row to prevent concurrent work. The claim lease is 6 minutes, longer than the scheduled collector route's 5-minute runtime limit; after a crash, the first timer tick after expiry reclaims it. Collection also uses a 10 minute overlap window and deduplication, encrypted raw evidence batches, and a monitoring-loss event when a feed has been failing for an hour. Stored events wait in a durable rule-evaluation queue and retry on a later run if evaluation fails. Drift checks, daily DNS checks (SPF, DKIM including both Microsoft DKIM target shapes, DMARC) and device-context mapping ride along.
  • Rules: three-valued logic (true, false, unknown) over the normalized events with baselines (a value never seen in N days); ten templates ship in monitor mode. Each edit publishes a new rule revision; matches reference the revision they were evaluated by.
  • Incidents: opened from matches, with transitions, notes, linked accounts and devices, and a SAAS alert on Monitoring > Alerts. Simulations replay a rule against stored history and save a simulation result, without opening real incidents or changing provider accounts.
  • Response: the implemented execution adapter is Microsoft Graph revokeSignInSessions, not a generic write API for every connected provider. A plan is built from the incident (for example revoke sessions for these Microsoft accounts), hashed, and executed only if the customer's response policy allows it at execution time; protected accounts are refused, a tampered plan is refused, an approval cannot be reused, and the kill switch stops even an approved plan. Configuration change plans are likewise Microsoft-specific (security defaults, SharePoint sharing and user-consent settings); their existence does not authorize equivalent changes to Slack, Google or the other providers.
  • Device context: cloud accounts are matched to managed devices through Intune, sign-in device names, the assigned user and the last logged-in user, with a confidence per method.

Step by step

  1. Open the customer’s Credentials & API > SaaS, select the intended provider and supply its indicated credential/authorization using the matrix below. Microsoft 365 collection is separate from Google Workspace’s built-in ITDR/import pipeline: Google has no collector in this normalized SaaS collection job. The connection appears under Security > SaaS Security > Integrations with its capability matrix.
  2. Detection rules: review the templates, enable the ones you want, choose monitor or alert per rule, and adjust thresholds. Use Simulate to see what a rule would have matched over the last days before switching it on.
  3. Response: set the customer's response policy (which actions are permitted, protected accounts, the approval requirement) and keep the kill switch where you can reach it.
  4. Posture and Hardening: review the findings and, where a change plan exists, review it before applying; nothing writes to the tenant without an explicit apply.
  5. Confirm the actual capability results, selected customer and successful feed samples; a catalog card or a stored credential is not successful collection. Re-check scopes and licensing with that customer’s administrator when refused.

Provider setup and coverage

The following are the credential recipes and feed descriptions implemented by the App, not independently tested acceptance for every vendor account. Use the callback shown by your workspace. For providers using a shared OAuth app (including Slack, Box, Webex, Egnyte and ShareFile), registration is shared across customers within that workspace unless deployment environment credentials take precedence; each customer still authorizes its own tenant. Do not paste client secrets into support tickets.

Dropbox

Team event log monitoring: logins, sharing, admin and file activity.

A Dropbox app (Scoped access) with the events.read and team_info.read team scopes, authorized by this client’s own team admin.

Citrix ShareFile

Activity report monitoring (field names are a best-effort read, not yet verified against a live tenant).

A Client ID and Client secret from an app registered at api.sharefile.com; the account admin authorizes it in a popup.

Box

Enterprise admin event monitoring: logins, sharing, admin role changes and file activity.

Shared Client ID and Client secret from a Server Authentication (Client Credentials Grant) custom app, App + Enterprise access and Manage enterprise properties scope, plus this customer's Enterprise ID. The customer's admin authorizes the shared Client ID in Box Admin Console > Apps > Custom Apps Manager; this UI has no Box OAuth popup.

Egnyte

Login audit report monitoring (field names are a best-effort read, not yet verified against a live tenant).

A domain, Client ID and Client secret from an app registered at developers.egnyte.com, authorized by this domain’s own admin.

Slack

Enterprise Grid audit log monitoring: admin actions, logins and workspace/app changes.

A Client ID and Client secret from a Slack app at api.slack.com/apps, with the auditlogs:read user scope and this org’s callback URL registered as a redirect URL.

Zoom

Sign-in/sign-out activity log monitoring.

An Account ID, Client ID and Client secret from a Server-to-Server OAuth app in the Zoom App Marketplace, created inside this client’s own Zoom account.

Webex

Admin Control Hub audit event monitoring.

A Client ID and Client secret from an Integration registered at developer.webex.com, with audit:events_read and spark:people_read scopes.

Salesforce

Login history and admin setup audit trail monitoring.

A My Domain hostname plus a Client ID (Consumer Key) and Client secret from a Connected App in Salesforce Setup, with the Client Credentials Flow enabled and a run-as integration user chosen.

Okta

System Log monitoring: sign-ins, MFA, admin and lifecycle events.

A domain and an SSWS API token from Admin Console → Security → API → Tokens, generated by a dedicated service account (the token stops working if its creating user is deactivated).

GitHub

Organization audit log monitoring (GitHub Enterprise Cloud organizations only).

An organization slug and a personal access token (classic) with the read:audit_log scope, generated by an organization owner.

1Password Business

Sign-in attempt and vault item usage monitoring.

An Events Reporting bearer token from 1Password.com → Integrations → Events Reporting → Add a token, scoped to sign-in attempts and item usages.

Atlassian (Jira/Confluence)

Organization event monitoring: sign-ins, group/policy changes and admin activity (Guard Premium or Enterprise plans only).

An organization ID (from the admin.atlassian.com URL) and an admin API key scoped to read:events:admin, created at admin.atlassian.com → Settings → API keys.

Duo Security

MFA authentication log monitoring: sign-in/push results, factor used and failure reasons.

A hostname, integration key and secret key from an "Admin API" application in the Duo dashboard (Applications → Admin API), granted "Grant read log" permission.

JumpCloud

Directory Insights monitoring: sign-ins, admin actions and directory changes (Directory Insights Premium only).

A per-admin API key from console.jumpcloud.com → admin menu → API Settings; MSP accounts also bring the target org ID.

OneLogin

Event monitoring: logins, MFA, admin actions and directory/app changes.

A subdomain, Client ID and Client secret from an API Credential in OneLogin (Developers → API Credentials), scoped to at least Read Users, authorized by this client’s own account owner or admin.

Keeper Security

Audit event (ARAM) monitoring: logins, sharing and vault activity (Enterprise + ARAM add-on only).

A region and a public API key minted once via the Keeper Commander CLI (public-api-key generate --roles "SIEM:1"); there is no admin-console page for this, only the CLI.

Bitwarden

Event monitoring: logins, admin actions, vault item and Secrets Manager activity (Teams/Enterprise on the US cloud only).

A Client ID and Client secret from Admin Console → Settings → Organization info → API key, generated by an organization owner. US cloud (bitwarden.com) only for now; EU-region and self-hosted organizations are not yet supported.

GitLab

Group audit event monitoring: members, settings and permission changes (Owner role required).

A host (gitlab.com or a self-managed instance), the numeric group ID, and a personal access token with the read_api scope, generated by a group Owner.

Egnyte/ShareFile report field parsing is explicitly best effort and unverified against a live tenant. Other providers also depend on actual plan, permissions, region and event availability. A capability marked available proves its bounded probe, not complete history or every event schema. Read last attempt/success, checkpoint, lag, partial/denied status and the actual sample before relying on coverage. Google identity monitoring belongs to ITDR; M365 permission setup is in Microsoft 365 management.

Verification

  • Integrations shows the tenant with feeds available and a recent successful collection. The collector is scheduled every 5 minutes; a feed is marked stale after 15 minutes without a successful run.
  • Check each feed's last attempt, last success, checkpoint, and lag. A failed feed keeps its last checkpoint and resumes from there after recovery.
  • Accounts lists the tenant's users with their last sign-in and device mappings.
  • For Microsoft session revocation, inspect each attempt and subsequent verification. Accepted or verification-pending is not proof that every session ended: already-issued access tokens can remain valid until expiry, and guests require handling in their home tenant. Revocation cannot be undone; it does not remove the separate need to investigate account access.

Providers, rule scope and sign-in details

The SaaS Security overview has a Providers table with one row per provider across the selected customers: customers connected, feed health with the first error, newest event, worst ingestion lag, events in the last day, sign-ins, failed sign-ins and events flagged as coming through a VPN, proxy, Tor or residential proxy. The event and sign-in counts link to that provider's events. Investigations and Incidents have a Provider filter and column, and the CSV export honors it.

A detection rule can be limited to the providers it was tuned for, so a download threshold set for SharePoint does not fire on Dropbox. The Detection rules list shows each rule's provider scope and an Alert or Monitor only switch, recorded as an audited revision. An incident in alert mode raises a SaaS source alert that follows the customer's notification, ticket and Cases routing.

Sign-ins from Microsoft 365, Dropbox, Citrix ShareFile and the other connected providers are enriched through the same geolocation keys ITDR uses: missing location is filled, the network operator is recorded, and a VPN, proxy, Tor or residential proxy verdict is shown on Investigations, with the service named where it is known.

Troubleshooting

  • A feed shows permission missing: for Microsoft 365, add the permission to the app registration and reconnect the tenant (guide 15); for a generic provider, re-check the provider-specific credential, permission and shared-app setup in the matrix and connection UI, then reconnect.
  • A feed is stale or failing: use Collect now, then read the feed's last error and suggested next step. If Collect now succeeds but several feeds later become stale together, check that the five-minute SaaS collector schedule is running. If a manual collection fails, check the Microsoft connection, client secret expiry, and the listed permissions.
  • Revoke sessions refused: the response policy does not permit it, the account is protected, or the plan hash no longer matches the incident; rebuild the plan.
  • An alert opens but no automatic response plan appears: the incident remains available for investigation, and the incident timeline records a failed response proposal when possible. Check the SaaS collector JobRun for the failure count, review the customer's response policy, and create a plan from the incident if appropriate.
  • A collection reports an incomplete feed: a page limit or unavailable audit blob made the result partial. Integrations marks the feed partial and shows the reason. Available events are retained, and the prior checkpoint is held when a blob could not be read so the next run retries that window. Review recurring partial results as a coverage or volume issue.