मुख्य सामग्री पर जाएं

Advanced SWG: Implement Threat Intelligence filtering

Implementation Effort: Low

User Impact: Medium

Overview​

In a Zero Trust security model, implicit trust is eliminated, and every access request is continuously validated based on identity, risk, and context. Threat Intelligence (TI) filtering is vital in this paradigm, as it enables dynamic, real-time protection against emerging internet threats—without assuming that any network traffic is inherently safe.

By integrating TI into Secure Web Gateway (SWG) policies, organizations can ensure that users and devices are rigorously evaluated before accessing resources. This process leverages up-to-date threat data from both Microsoft and multiple third-party sources to identify and block connections to destinations classified as "High Severity" risks—such as phishing sites, malware distribution points, botnets, and command-and-control servers—at the FQDN or URL level.

Key capabilities:

  • Identify and block access to destinations labeled as "High Severity" threats based on FQDN or URL (threats identified according to Microsoft and 3rd Party Threat intel sources)
  • Configure allow list of FQDNs to handle false positives
  • Enrich the traffic log with the Threat Type (e.g., Botnet, C2, Phishing, Malware, etc.) of the matched indicator
  • Provide user and context-aware policy through identity-centric Conditional Access policy integration

Implementation steps:

  • Enable Internet Access traffic forwarding profile and assign users/groups
  • Create threat intelligence policy (automatically includes rule blocking high severity threats)
  • Configure allow list rules for known false positives or business-critical sites (optional)
  • Link threat intelligence policy to security profile (or use baseline security profile for tenant-wide protection)
  • Create Conditional Access policy targeting "All internet resources with Global Secure Access" and assign security profile
  • Verify enforcement by testing access to the documented safe test destination (entratestthreat.com) and reviewing Traffic logs for Threat Type metadata

Verify Threat-Intelligence Filtering for AI and Agent Destinations​

Implementation Effort: Low - Reuse the network baseline where available, or configure and test a threat-intelligence policy.
User Impact: Low - Malicious destinations are blocked; false positives can affect legitimate integrations.
Documentation checked: 2026-11-11
Ownership: Shared network baseline checkpoint; do not duplicate an existing equivalent control.

Overview​

AI applications and agents can follow attacker-controlled links or invoke compromised external services. Global Secure Access (GSA) threat-intelligence policies block supported traffic to known malicious destinations using threat indicators.

This workshop already references threat-intelligence filtering as a shared network capability. Verify that acquired AI traffic is covered by the existing protection; a reference alone does not demonstrate enforcement.

This supports Assume breach by limiting known malicious outbound communications and Verify explicitly through observable network decisions.

Prerequisites​

  • Complete AI_163 and establish acquisition for the source.
  • Internet Access licensing and appropriate GSA/Conditional Access administrative roles.
  • A threat-policy owner and a controlled validation window.

Implementation guidance​

  1. Review the existing threat-intelligence policy with the network team. Reuse it if it covers the intended AI traffic.
  2. If missing, create the threat-intelligence policy using the linked configuration guide and its supported indicator/action options.
  3. Link it through the appropriate security profile or baseline. Use AI_027 for the supported Copilot Studio path.
  4. Enable TLS inspection where HTTPS URL indicators must be evaluated. Domain-level protection and URL payload visibility have different requirements.
  5. Validate with a test destination provided in our configuration guide, such as entratestthreat.com, in an approved test environment. Do not use a live malicious destination.
  6. Confirm GSA is the enforcement point by correlating logs; an endpoint/browser block alone does not prove the GSA policy worked.
  7. Establish false-positive review and a controlled exception process.

Validation and completion evidence​

  • A documented safe test destination is blocked by GSA with the expected threat information in logs.
  • A legitimate AI destination remains available.
  • The policy/profile and traffic-source coverage are recorded.
  • Exceptions, propagation, and review ownership are documented.

Scope and limitations​

Threat intelligence covers known malicious destinations, not every unsanctioned AI service, novel prompt injection, sensitive prompt, or dangerous tool. Combine it with AI app filtering, MCP restrictions, and network DLP.

If another security layer prevents the test, do not disable production defenses to force it through. Use isolated testing and independently verify the intended network enforcement.

Reference​