Hunt with the Microsoft Sentinel platform
Hunting queries, KQL jobs in the data lake, summary rules, and notebooks with the Sentinel MCP server.
- 16 min
- 4 concepts
- Try it yourself
- 4-question check
Learn
Sentinel hunting covers both recent data in the analytics tier and years of history in the data lake. Jobs and summary rules bring the right slice of lake data back to where it can be queried fast.
- 1
Create and monitor hunting queries
Sentinel ships hunting queries from content hub solutions mapped to MITRE ATT&CK. Run them, save your own, add results to a hunt (a container for a hypothesis-driven investigation with notes and bookmarks), and use livestream to watch for new matches. Bookmarks save interesting rows and can be promoted to incidents.
Remember: Bookmarks keep evidence; hunts organise the work.
- 2
Create and manage KQL jobs in Data lake
KQL jobs run a KQL query against data lake tier data, once or on a schedule, and write the results to a table in the analytics tier. Use them to promote just the relevant subset of cheap long-term data (for example a year of sign-ins for one suspicious IP) so it can be hunted quickly or used by detections.
Remember: Lake → analytics: a KQL job.
- 3
Create and manage Summary rule tables for querying
Summary rules run scheduled KQL aggregations over high-volume data (including Basic and Auxiliary or lake data) and write compact results to a summary table in the analytics tier. Detections and workbooks then query the small summary table instead of the raw logs, cutting cost.
Remember: Aggregate noisy logs into a small table.
- 4
Hunt for threats by using Notebooks, including connection to the Sentinel MCP Server
Jupyter notebooks (in VS Code with the Microsoft Sentinel extension) run Python and Spark against data lake tables for advanced analytics, machine learning and visualisation, and can be scheduled as jobs. The Sentinel MCP server exposes Sentinel tools (such as data exploration and table search) to AI agents and assistants over the Model Context Protocol, so you can explore lake data with natural language.
Remember: Big analytics and ML on lake data: notebooks.
Hunting across tiers
- HypothesisFor example: a leaked key was used months ago.
- Explore the lakeNotebooks or KQL across years of data.
- PromoteA KQL job copies matches to the analytics tier.
- Hunt fastHunting queries, bookmarks, livestream.
- DetectTurn findings into analytics rules.
Try it
Which Sentinel hunting feature?
This interactive part needs JavaScript.
In the portal
- HuntingMicrosoft Defender portal: Microsoft Sentinel > Threat management > Hunting
- Data lake explorationMicrosoft Defender portal: Microsoft Sentinel > Data lake exploration > KQL queries
- JobsMicrosoft Defender portal: Microsoft Sentinel > Data lake exploration > Jobs
- Summary rulesMicrosoft Defender portal: Microsoft Sentinel > Configuration > Summary rules
Key terms
Click a card to see the definition.
Check your understanding
4 questions. Score 80% or more and these skills are marked as confident in your SC-200 study plan.
What moves data from the data lake to the analytics tier?
Jobs write lake query results to analytics tables.
What reduces high-volume logs to small aggregated tables on a schedule?
Summary rules aggregate data.
Where do you save interesting hunting results as evidence?
Bookmarks can be promoted to incidents.
Which tool runs Python and Spark against lake data?
Jupyter notebooks.
Sources
- Microsoft Learn: Threat hunting in Microsoft Sentinel
- Microsoft Learn: Microsoft Sentinel data lake overview
- Microsoft Learn: Aggregate data with summary rules
- Microsoft Learn: Explore the data lake with the Sentinel MCP server
Facts in this lesson are from Microsoft Learn. This site isn't affiliated with Microsoft.