ThreatScope

GitLab, SonicWall, Atlassian & AI Infrastructure Security Risks | October 7–13, 2026

Reporting period: October 7–13, 2026 Threat level: High for affected and exposed systems

🎧 Listen to this week’s ThreatScope (audio brief)

Executive Snapshot

This week’s ThreatScope focuses on vulnerabilities affecting software development platforms, remote-access gateways and enterprise collaboration applications, alongside a malware campaign targeting exposed artificial intelligence (AI) infrastructure.

The highest-severity finding is GitLab CVE-2026-97229, which can allow unauthenticated server-side code execution through specially crafted continuous integration and continuous delivery (CI/CD) configuration. The vulnerability is confirmed by the vendor, although exploitation has not been established in the reviewed advisory.

SonicWall SMA1000 and Atlassian Data Center also require urgent attention following independent reports of exploitation attempts. These observations indicate attacker interest but do not establish successful compromise.

An additional concern is the PoeLLM campaign, which demonstrates how exposed AI and application services can be compromised and reused for cryptocurrency mining, remote access and further attacks.

Is This Report Relevant to Your Organization?

This report is particularly relevant if your organisation operates:

Prioritisation should consider the affected version, external reachability, available privileges, exposure period and access to connected systems.

Executive Summary

Four vulnerabilities and one active malware campaign deserve attention this week.

GitLab CE/EE CVE-2026-97229, rated CVSS 10.0, affects the handling of regular expressions in CI/CD configuration. Under certain conditions, a specially crafted expression can lead to integer overflow and unauthenticated code execution on the server. GitLab has released corrected versions, but the reviewed advisory does not establish active exploitation.

SonicWall SMA1000 CVE-2026-102255, with a vendor severity score of 10.0, affects the Work Place interface. The vulnerability permits server-side request forgery (SSRF) before authentication. SonicWall initially reported no evidence of exploitation, while subsequent independent monitoring identified exploitation attempts.

Atlassian CVE-2026-21589, rated CVSS 4.0 at 9.3, affects multiple Data Center products. An unauthenticated attacker who knows the exact file path and name may access certain files within the application directory. Independent monitoring has reported exploitation attempts.

Citrix NetScaler CVE-2026-88779 remains relevant from the previous reporting period. In affected configurations using Security Assertion Markup Language (SAML), the vulnerability can cause denial of service. It should not be confused with separate NetScaler vulnerabilities that allow remote code execution.

Finally, PoeLLM / Canto Incognito is a malware campaign targeting exposed AI and application services. Lumen’s research describes remote-shell functionality, cryptocurrency mining and the reuse of compromised servers for further scanning and exploitation.

The common operational concern is the potential for an initial vulnerability or exposed service to provide access to additional systems, credentials or infrastructure.

Key Vulnerabilities

Vulnerability Analysis

GitLab CE/EE — CVE-2026-97229

Status: Vendor-confirmed vulnerability; exploitation not established CVSS 3.1: 10.0

Insufficient validation of regular expressions in GitLab’s CI/CD processing can result in integer overflow. Under certain conditions, an unauthenticated attacker may exploit a specially crafted expression to execute code on the GitLab server.

The vulnerability is particularly significant because development platforms may hold deployment credentials, access tokens, application secrets and connections to production infrastructure.

GitLab identifies the following corrected versions:

The vendor’s corrected version for the 19.3 branch is 19.3.5.

The security update also addresses other issues, including protected-branch pipeline authorisation. Organisations should account for database migrations and potential downtime when planning upgrades.

Required Response

Prioritise externally accessible, self-managed GitLab installations and deploy the appropriate corrected version.

Before disruptive maintenance, preserve application logs, authentication records and CI/CD audit events.

Investigate unexpected changes to pipeline definitions, protected branches, runners and access tokens, together with unusual processes or outbound connections from GitLab hosts.

Where compromise is suspected, assess whether deployment credentials, generated software artefacts or connected cloud environments may also have been exposed.

The investigation should reflect the permissions available to the affected GitLab service.

SonicWall SMA1000 — CVE-2026-102255

Status: Independent exploitation attempts reported Vendor severity score: 10.0

CVE-2026-102255 affects the SonicWall SMA1000 Work Place interface.

The vulnerability allows an unauthenticated attacker to induce server-side requests through an unintended access path. This creates a risk of unauthorised interaction with internal functionality.

SonicWall identifies the following corrections:

SonicWall’s October 6 notice stated that it had no evidence of exploitation. Subsequent reporting from Previdian identified attempts observed through independent monitoring.

This distinction matters: observed exploitation attempts demonstrate attacker activity but do not independently confirm successful compromise of a production system.

Required Response

Apply the appropriate correction urgently and verify the actual running build on every appliance, including standby systems.

Preserve appliance logs and relevant network records. Review unexpected outbound requests, unusual sessions and administrative or configuration changes.

Confirm that an earlier security correction has not been mistaken for the October update addressing this vulnerability.

Atlassian Data Center — CVE-2026-21589

Status: Independent exploitation attempts reported CVSS 4.0: 9.3

CVE-2026-21589 affects several Atlassian Data Center applications.

The vulnerability allows an unauthenticated attacker to access certain files within the web application’s root directory.

Exploitation requires knowledge of the exact filename and path. The vulnerability does not independently provide a way to enumerate directories.

Affected products include Bitbucket, Confluence, Jira, Bamboo, Crowd, Crucible and Fisheye.

Selected vendor-listed corrected versions include:

Atlassian states that affected Cloud products have been patched. This should not be interpreted as evidence that customer-managed Data Center deployments were unaffected.

Independent monitoring has identified exploitation attempts, although individual requests do not establish successful file access.

Required Response

Upgrade affected applications, including cluster nodes and applicable mirrors.

Where immediate upgrading is not possible, restrict external access or apply documented vendor mitigations.

Review access logs for suspicious file-access requests, including encoded paths. Determine whether sensitive configuration files or credentials were stored in potentially accessible locations.

Rotate credentials where investigation supports possible exposure.

Citrix NetScaler — CVE-2026-88779

Status: Continuing exposure from the previous reporting period CVSS 4.0: 8.7

CVE-2026-88779 is a memory-overflow vulnerability affecting NetScaler ADC and Gateway when configured as a SAML service provider or identity provider.

Successful exploitation may cause denial of service and disrupt authentication-dependent services.

The vendor identifies these corrected versions:

• 14.1-73.41 or later

• 13.1-64.28 or later

• 14.1-FIPS 14.1-73.41 or later

• 13.1-FIPS / NDcPP 13.1-37.282 or later

The documented impact is service disruption, not remote code execution.

Required Response

Verify whether the affected SAML configuration is enabled and confirm the installed software version.

Review unexpected crashes, authentication failures and service interruptions.

Validate failover procedures and emergency administrative access, particularly where NetScaler supports critical business applications.

AI Infrastructure Threat Watch — PoeLLM / Canto Incognito

Status: Observed malware campaign

Research published by Lumen’s Black Lotus Labs describes a campaign targeting exposed AI and application services.

The campaign, active since April, involves compromised servers used for remote-shell access, cryptocurrency mining, additional scanning and exploitation.

The research identifies LiteLLM CVE-2026-42271 as a likely exploitation path in an analysed sample.

Other services, including Ollama, Gitea and Gotenberg, were observed in infrastructure associated with victims. This does not establish that each product was compromised through a specific vulnerability.

The malware also retrieves command-and-control addresses from text hosted on GitHub, allowing operators to change their infrastructure without replacing every malicious component.

Required Response

Include AI gateways, model-serving systems and development utilities in the organisation’s asset inventory.

Remove unnecessary public access and update affected applications.

Investigate unexplained processing demand, cryptocurrency-mining activity, unexpected remote shells and outbound scanning.

Where compromise is confirmed, rebuild affected workloads from trusted sources and rotate accessible service credentials.

For AI environments, the investigation should also consider the permissions of service accounts and access to connected applications, data and development infrastructure.

What Deserves Attention This Week?

Recommended Management Actions

Immediate Actions: 0–7 Days

• Identify affected GitLab, SonicWall, Atlassian and Citrix systems.

• Inventory externally exposed AI and application services.

• Prioritise emergency updates for exposed GitLab and SonicWall installations.

• Upgrade or contain affected Atlassian Data Center applications.

• Verify NetScaler SAML configurations and corrected versions.

• Preserve relevant evidence before disruptive remediation.

• Investigate suspicious development activity, gateway requests and file-access attempts.

• Assess unexpected processing demand, remote shells and outbound scanning in AI environments.

• Review potentially exposed credentials and connected systems.

• Document systems where available evidence is insufficient to exclude compromise.

Actions Within 30 Days

• Independently verify corrected versions across all affected installations and cluster nodes.

• Complete documented compromise assessments for exposed development, remote-access and collaboration systems.

• Review GitLab deployment credentials, runners and trusted integrations.

• Determine whether Atlassian configuration files or secrets were accessible.

• Verify that SonicWall standby appliances received the required corrections.

• Review logging coverage and retention for affected systems.

• Improve inventory of AI services and development utilities.

• Validate recovery procedures for critical applications and access infrastructure.

• Assign owners to unresolved exposure and investigation findings.

Strategic Actions: Within 90 Days

• Reduce unnecessary internet exposure of development and administrative interfaces.

• Include AI workloads and externally managed applications in vulnerability-management processes.

• Strengthen privileged access controls around development and remote-access platforms.

• Improve centralised monitoring of application, network and system activity.

• Review how deployment credentials and service secrets are stored and accessed.

• Establish monitoring for unexpected computing-resource consumption and outbound scanning.

• Validate recovery procedures for development platforms and identity-dependent services.

• Define closure criteria that distinguish technical remediation, compromise assessment and recovery validation.

• Maintain documented accountability for unresolved risks and incomplete evidence.

Enterprise Exposure Assessment

Four questions should guide prioritisation this week.

1. Was the affected system externally reachable?

Check internet exposure, administrative interfaces, access restrictions and historical network configuration.

2. What could successful exploitation provide?

GitLab code execution, SonicWall server-side requests and Atlassian file access have different consequences and require different investigation approaches.

3. Which connected systems or sensitive resources could be affected?

Consider deployment credentials, application configuration, internal services, identity systems and cloud environments.

4. Is there sufficient evidence to assess previous compromise?

Preserve relevant logs and investigate activity during the plausible exposure period.

Observed exploitation attempts should trigger appropriate investigation, but should not automatically be treated as proof of successful compromise.

Missing records should be documented rather than interpreted as evidence that no compromise occurred.

Questions for Leadership

Security and service owners should establish:

• Are any affected GitLab or SonicWall systems accessible from untrusted networks?

• Have actual installed versions been verified?

• Are all affected Atlassian Data Center applications included in remediation?

• Could exposed application files contain credentials or sensitive configuration?

• Do GitLab services have access to production deployment infrastructure or cloud environments?

• Are AI workloads and supporting services included in the asset inventory?

• Has unusual processing demand or outbound scanning been investigated?

• Are critical NetScaler SAML services running corrected versions?

• Have potentially compromised systems been investigated before recovery?

• Are remediation and compromise assessment tracked as separate decisions?

A finding should be closed only when the required correction or mitigation has been verified, the investigation outcome is documented and any necessary recovery actions are complete.

ISO 27001 & ISO 9001 Alignment

The recommended actions support established security and quality-management practices, including:

• asset and vulnerability management;

• secure configuration and change management;

• identity and access control;

• logging and monitoring;

• incident investigation and response;

• protection of credentials and sensitive information;

• supplier and externally managed service oversight;

• business continuity and recovery;

• corrective action and independent verification;

• documented accountability and residual-risk management.

For exposed systems, evidence should cover both the technical correction and the outcome of any necessary compromise investigation.

Need Help Assessing Your Exposure?

If your organisation uses the technologies covered this week, begin by confirming the actual running versions, external reachability and potential access to connected systems.

Where suspicious activity is identified, the investigation should extend to application changes, credentials, deployment processes, network activity and available historical evidence.

DIAMATIX supports organisations through:

• Security Operations Center as a Service (SOCaaS)

• Managed Detection and Response as a Service (MDRaaS)

• Vulnerability Management

• Incident Response

• Digital Forensics

Contact DIAMATIX to discuss your environment.

Conclusion

This week’s findings demonstrate how vulnerabilities in development platforms, remote-access gateways and enterprise applications can create risks beyond the initially affected system.

For GitLab, the priority is to correct a critical code-execution vulnerability and assess the potential exposure of development and deployment infrastructure.

For SonicWall SMA1000, independent reports of exploitation attempts justify urgent remediation and investigation, while remaining distinct from confirmed successful compromise.

For Atlassian Data Center, the focus is on correcting affected installations and determining whether sensitive application files may have been accessed.

For Citrix NetScaler, the continuing concern is service availability in affected SAML configurations.

The PoeLLM campaign also reinforces the importance of securing AI workloads and the supporting application services that may be accessible from the internet.

For affected and exposed systems, the practical response sequence remains:

Identify Exposure → Contain → Remediate → Assess for Compromise → Recover → Validate

The objective is to establish whether the vulnerability has been corrected, whether previous compromise has been adequately investigated and whether the affected service can be trusted again.

Sources

• ThreatScope by DIAMATIX weekly research

• GitLab Security Releases and Documentation

• SonicWall Security Advisory

• Atlassian Security Advisory

• CERT-EU

• Citrix Security Advisory

• Previdian Security Research

• Lumen Black Lotus Labs

Methodology

ThreatScope by DIAMATIX prioritises vulnerabilities and threats according to exploitation evidence, system reachability, available privileges and potential business impact.

Reported exploitation attempts are distinguished from confirmed successful compromise. Remediation and compromise assessment are evaluated separately.

Trusted · Innovative · Vigilant