AuditTeam: who can stop your endpoint protection?
The AuditTeam case calls for checking who can change endpoint protection. Connect permissions, applied policy and evidence for accepting recovery.
Ler em portuguêsIn this article
By Pentesty
Start with a decision your team can substantiate
In its November 18, 2026 announcement, Kaspersky attributes to AuditTeam a campaign observed from January through October across at least seven countries and territories. It describes OpsAgent with one module using a signed driver to terminate processes and another encrypting files. Likely AI assistance is the researchers’ assessment. These observations do not establish a compromise of your organization.
The demand presented as an audit is extortion. For an organization, the useful question is which permissions and changes could allow defenses to be interrupted. Use the technical report to guide investigation while separating vendor observations, evidence found in your environment and unresolved hypotheses. Each conclusion should identify its source, time and the person responsible for verification.
Assign responsibility for every privileged change
Choose a priority service and identify who administers its computers, protection tools and support access. For a driver installation, require a verifiable request, purpose, package origin, cryptographic identifier, approver and maintenance window. Check that the account used matches the authorization and that the record allows the change to be reconstructed afterwards.
A request made under pressure needs verification through an independently established contact. The FaceStealer analysis helps design that approval path. In an exercise, present a fictional request and check whether support and operations can validate the responsible person before granting access. Keep the result and the exceptions discovered.
Check what the driver policy actually covers
Microsoft’s ASR reference distinguishes preventing vulnerable drivers from being written to disk from preventing existing files from loading. Its driver blocking guidance recommends validating policies in audit mode before enforcing them.
Turn that distinction into an applicability test: which Windows version, policy and devices are included in the change? On a representative sample, record the effective configuration and the corresponding policy event. Coordinate compatibility assessment with the service owner. A blocklist being present, by itself, does not establish coverage of the driver described in this campaign.
- Inventory: device, role, operating system version and owner.
- Change: approved package, identified policy and application time.
- Verification: observed event, sample coverage and limitations.
- Continuity: legitimate function tested, result and rollback condition.
Prepare to respond when visibility disappears
Decide who checks a connected computer when its agent disappears from the console. Compare endpoint signals with inventory, network information and authorized changes before concluding what caused it. Send discrepancies to the incident response owner and preserve available records in access-controlled storage. Also record collection gaps and clock differences that complicate correlation.
Use benign, previously approved signals in a laboratory to exercise this workflow. The ArmorSteal analysis complements checks of protection settings and access associated with an endpoint. The exercise should show who receives the alert, which evidence they consult and which decision they can make, with defined deadlines and owners.
Accept recovery one service at a time
Before restoring access, establish what must be trustworthy: protection settings, administrative accounts, sessions and data needed for operations. Separate evidence of containment, investigation and rebuilding. For a backup, request a controlled restoration and the result of a realistic function using synthetic data. Include dependencies and permissions, beyond reading a restored file.
The Hafnium analysis organizes similar questions about returning to operation. Here, keep a record for each service with its observed state, acceptance owner and outstanding work. An exception needs a deadline and a review condition. Collect this material so another team can examine the decision to resume activity.
Frequently asked questions
Does a signed driver authorize its installation?
Authorization must connect the request, responsible person and device. Verify the package’s origin and approval even when it carries a digital signature.
Does a blocklist complete the assessment?
Verify coverage, effective configuration and results on the Windows systems assessed. Record compatibility and exceptions; the list does not replace investigation of earlier activity.
Did Kaspersky prove AI was used?
Researchers describe signs consistent with AI-assisted development. This is an assessment, not proof of the provenance of all the code.
What should support recovery acceptance?
Collect protection settings, access review, restoration results and service continuity evidence. Identify coverage, limitations and who approved each result.
Sources and references
- Kaspersky Securelist: OpsAgent and AuditTeam technical research, November 18, 2026
- Kaspersky: AuditTeam campaign announcement, November 18, 2026
- Microsoft: recommended driver block rules
- Microsoft: scope of the vulnerable signed driver ASR rule
- CybersecurityNews: AuditTeam editorial reference, November 19, 2026
Original Pentesty analysis, published November 19, 2026. The proposed scenarios are editorial guidance, with scope and validation to be defined for each environment.