Muistiinpano
Tämän sivun käyttö edellyttää valtuutusta. Voit yrittää kirjautua sisään tai vaihtaa hakemistoa.
Tämän sivun käyttö edellyttää valtuutusta. Voit yrittää vaihtaa hakemistoa.
Use the Recon Agent in Microsoft Perception to make complex attack-path analysis easier and faster to understand. The agent reasons over relationships between Azure resources, Microsoft Entra ID identities, and permissions to explain potential attack paths. Its findings and supporting evidence help you understand how a path could work, without executing attacks or changing resources.
Choose a playbook based on your investigation goal, run a session in an authorized scope, and review the findings to decide what to investigate or address next.
For prerequisites, setup, and session management, see Set up and manage the Recon Agent.
Why use the Recon Agent?
Microsoft Security Exposure Management provides ongoing visibility into security posture and helps you prioritize exposure risk across your organization.
The Recon Agent complements exposure management with focused reasoning and evidence-backed explanations of potential attack paths. It helps you understand how individual permissions and resource relationships could combine into a path to critical assets, rather than assessing each finding in isolation. Each session provides a read-only, point-in-time assessment based on live data. Use the explanations and supporting evidence to decide what to investigate or address next.
For example, when you're investigating a potentially compromised user, use Recon findings to understand how that user's permissions and relationships with other identities or Azure resources could create a path to critical assets. Review the evidence for the reported steps with the relevant resource and identity owners before deciding on follow-up.
The Recon Agent provides the following capabilities:
- Attack-path reasoning: Uses data from your Azure and Microsoft Entra ID environment to analyze how identities, permissions, and resources could combine into potential attack paths.
- Evidence-backed explanations: Explains the relationships behind reported paths so you can understand how a path could work and review the supporting evidence.
- Multi-agent workflows: Participates in Protect against a threat alongside the Threat Intelligence, Posture Prioritization, and Detection Authoring agents.
Note
The Recon Agent creates a read-only assessment. It doesn't exploit vulnerabilities, execute payloads, modify resources, deploy detections, persist attack graphs, or operate outside your organization. Results are a point-in-time snapshot and don't replace continuous monitoring systems.
Choose a Recon use case
Choose a playbook that matches your goal and the inputs you have available.
| Goal | Playbook | Required input |
|---|---|---|
| Investigate a threat described in a threat intelligence article. | Protect against a threat | A threat intelligence article. |
| Review risks associated with a managed identity. | Assess identity risks | An Azure subscription and managed identity. |
| Investigate potential attack paths in an Azure subscription. | Identify attack paths | An Azure subscription. |
Operate the Recon Agent
Before you start, define the question you want to answer, confirm that you're authorized to assess the target environment, and identify who will review the findings and coordinate follow-up.
Assign access
Have a user with Security Admin set up and configure the Recon Agent. Users with Security Reader or Security Admin can start, view, and interact with agent sessions after setup.
The agent needs read-only Microsoft Graph application permissions and Azure roles assigned with the permission assignment script. Granting Microsoft Graph permissions requires Privileged Role Administrator in Microsoft Entra ID. Assigning Azure roles requires Role Based Access Control Administrator on each target subscription.
For role requirements, see Required permissions and Agent permissions. Complete agent setup with PowerShell and assign user access before those users start sessions.
Respond to a new CVE or threat
Run Protect against a threat when you're concerned that a newly published Common Vulnerabilities and Exposures (CVE) record or a known threat could affect your environment. The playbook provides a relevant threat intelligence article as input, then review the playbook findings to decide what needs investigation or follow-up.
Investigate managed identity access to critical assets
Run Assess identity risks when you're investigating whether a managed identity could provide a path to critical assets, sometimes called crown jewels. Select the Azure subscription and managed identity, then review the reported paths and supporting evidence against the assets your organization considers critical.
Investigate attack paths in a subscription
Run Identify attack paths to investigate potential attack paths in an Azure subscription. Select an authorized subscription, then review the reported paths and supporting evidence.
For session entry points and the managed-identity walkthrough, see Start a Recon Agent session. After reviewing the results, use Act on the findings to plan follow-up.
View Recon Agent output
The session displays Recon Agent output as Markdown files.
Act on the findings
Use reviewed findings to choose the next action in your existing security workflow:
- Investigate further. If evidence is incomplete or a claim is uncertain, gather additional information before treating the finding as confirmed.
- Address confirmed issues. Work with the responsible resource or identity owner to evaluate changes through your normal change-management process. Recon doesn't apply remediation.
- Validate a potential path. If you need to test whether a path can be exploited, use a separately authorized security-testing process. Recon doesn't perform exploitation.
Repeat an assessment after relevant changes when you need to revisit the original question. For recurring assessments, choose a cadence based on changes in the target environment and your team's capacity to review and act on the results.
A run with no findings doesn't establish that the environment is secure. Interpret the result within the scope, permissions, and data available to that session.