How to Detect Impossible Travel in Microsoft 365 Sign-Ins

Summary
Huntress explains a six-step approach to detecting impossible travel in Microsoft 365 sign-ins and reducing false positives from legitimate activity, high event volumes, and inaccurate geolocation data.
Key points
- Impossible-travel detection flags a user account appearing in geographically distant locations within a timeframe too short for normal travel.
- The article focuses on Microsoft 365 UserLoggedIn events, which indicate successful authentication and include contextual information.
- A six-step pipeline pairs events, calculates implied velocity, filters possible travel and normal behavior, assigns risk, and excludes low-risk pairs.
- Same-timestamp or near-simultaneous logins can result from services or devices authenticating to one account, or shifts between private networks and mobile devices.
- To manage high data volumes, prioritize relevant logs, aggregate recent user and location events, and defer expensive processing.
- User baselines and additional event context help distinguish benign legitimate activity from suspicious anomalies.
- Because geolocation data can be inaccurate, the article recommends accounting for distance and aggregating results from multiple location databases.
Article Details
- Defense Focus
- Detect potential user-account compromise by identifying successful logins from geographically distant locations within an implausibly short time, then filtering and prioritizing anomalies.
- Detection Methods
- Pair user events by a user attribute such as user ID or email address.
- Calculate the time difference and implied velocity between event pairs, applying a predetermined threshold.
- Exclude pairs consistent with possible travel, then compare remaining anomalies with the user's baseline behavior.
- Assign risk scores using predetermined criteria and filter low-risk pairs, including possible credential sharing.
- Improve geolocation confidence by aggregating results across multiple location databases; treat longer distances as stronger velocity evidence.
- Data Sources
- Microsoft 365 Unified Audit Log, particularly UserLoggedIn events.
- User events with timestamps, location attributes, and user identities.
- Platforms
- Microsoft 365
- Defensive Actions
- Prioritize relevant logs and aggregate around the most recent user and location events to manage high data volume.
- Place more computationally expensive operations later in the detection pipeline.
- Use user baselines and contextual enrichment to distinguish malicious activity from legitimate usage.
- Account for legitimate VPN or proxy activity, credential sharing, multiple devices, and inaccurate geolocation data when triaging alerts.
