Active in the Wild: The 10 Exploited Vulnerabilities That Matter Most | Nov. 15–21, 2026

An actively exploited vulnerability becomes strategically important when it gives an attacker authority that extends beyond the vulnerable host. This week’s lead example is VMware vCenter CVE-2026-123890: an unauthenticated authentication bypass used before a fix existed to create vSphere SSO administrator access, obtain identity material and support lateral movement.

 

That is the distinction this ranking is designed to make. It does not sort by CVSS or advisory date. It weighs reachable attack surface, attack prerequisites, authority gained, deployment prevalence, exploit evidence, remediation friction and the business systems behind the product.

 

Eight entries are new to this weekly edition. Two—Barracuda ESG and Check Point DynamicCLI—remain continuing active threats because they still represent credible edge-to-root paths that must be verified closed. VMware vCenter is the only entry labeled ZERO-DAY because public forensic evidence establishes exploitation before the November 19 patch.

 

Executive Takeaways

 

  • Control planes lead the week. vCenter, Puppet and security gateways can turn one foothold into administrative reach across many systems.

  • Internet-facing and boundary systems remain disproportionately valuable: Cisco Expressway, Barracuda ESG, Apache Roller and public REDCap survey flows all cross external trust boundaries.

  • Patch age is not remediation evidence. SAP and Synology fixes were available before this week, yet new KEV confirmation shows unresolved exposure still exists.

  • Legacy software creates an ownership problem as much as a patch problem. Apache Axis 1.x is end-of-life and may be embedded inside applications that no team inventories as “Axis.”

  • Verification must extend beyond version checks when compromise may predate the fix. Identity, token, root-level and configuration persistence require separate review.

 

Ranked Top 10

 

1. VMware vCenter Authentication Bypass — CVE-2026-123890

NEW THIS WEEK | ZERO-DAY | Confirmed exploitation

 

Broadcom describes an authentication bypass in VMware Directory Service. Mandiant and Google Threat Intelligence Group add the operational evidence: ARMORY CASTLE exploited the flaw in early September, while Broadcom’s fixes arrived November 19. Their observed BONESAW tooling could create an [email protected] session, extract the vCenter LMDB, obtain the SAML STS signing key, forge assertions and create administrator accounts.

 

Attack path: Network access to the vCenter DCE/RPC service → SRP session-key disclosure → vSphere SSO administrator session → directory/database access, SAML trust compromise and lateral movement.

 

Why it matters: vCenter administers ESXi hosts and virtual machines. Its identity material can support persistence outside the original exploit transaction, and internal-only reachability does not remove risk from an actor that already has a foothold.

 

Remediation and validation: Apply the fixed vCenter release for every supported branch. Segment management services, investigate vmdird RPC activity and LMDB access, inventory recent SSO and machine-account creation, review STS key access and assess whether token-signing material must be rotated. Verify the installed build from the appliance, not only the change ticket.

 

2. Cisco Expressway Hard-Coded Credential RCE — CVE-2026-76466

NEW THIS WEEK | Confirmed exploitation

 

CISA added the vulnerability to KEV on November 20. Cisco says an unauthenticated remote attacker with network access and persistent cookies can invoke database-service RPC commands on a clustered Expressway device and execute commands as root. Fixed software is available; Cisco provides no workaround.

 

Attack path: Reachable Expressway database service + possession of persistent cookies → RPC invocation → root command execution on Expressway-C or Expressway-E.

 

Why it matters: Expressway is often deployed at a communications boundary and connected to collaboration and internal identity services. Root on that device can create both persistence and a trusted pivot point.

 

Remediation and validation: Upgrade to a fixed version, restrict service reachability, verify cluster and client-certificate behavior, rotate exposed persistent material where the incident scope requires it, and review root-level process, file and connection activity. Because no workaround exists, an unpatched exposed device requires a containment decision, not a monitoring-only plan.

 

3. Barracuda ESG Pre-Authentication Authorization Bypass — CVE-2026-116606

CONTINUING ACTIVE THREAT | Vendor-observed exploitation

 

Barracuda states that exploitation affected a small number of organizations. Authentication is not required, but the administrative UI must be reachable. On ESG, the flaw can be paired with CVE-2026-111560 command injection to achieve root access.

 

Attack path: Reachable administrative UI → unauthorized configuration-import capability → companion command-injection path → root authority.

 

Why it matters: Email gateways handle high-trust traffic, credentials and policy. Administrative exposure collapses the intended separation between perimeter inspection and device control.

 

Remediation and validation: Apply current firmware and patches, restrict the Admin UI to dedicated management networks, validate patch presence on-device, and investigate configuration imports, administrator events and command execution. If the device may have been compromised, rebuilding or replacing it can be more reliable than assuming patching removed persistence.

 

4. Apache Roller Unauthenticated XML-RPC Deserialization RCE — CVE-2026-129448

NEW THIS WEEK | Apache-confirmed active exploitation

 

Apache says Roller 4.0.0 through 6.1.4 can deserialize vendor-extension types at the XML-RPC endpoint before authentication and that the issue is actively exploited. Upgrade guidance points to 6.1.6 or later. Disabling XML-RPC in the administration UI alone is not an effective mitigation for affected older versions because parsing occurs first.

 

Attack path: HTTP request to /roller-services/xmlrpc → pre-authentication deserialization → arbitrary code execution.

 

Why it matters: A control may appear disabled while the vulnerable processing path remains reachable. That is a verification failure as much as a software defect.

 

Remediation and validation: Upgrade. If an interim measure is required, disable extension types or remove the servlet mapping exactly as Apache directs, restart Roller, and retest the endpoint externally. Hunt for unusual XML-RPC bodies, Java process anomalies, child shells and unexpected outbound connections.

 

5. Check Point DynamicCLI Local Privilege Escalation — CVE-2026-123039

CONTINUING ACTIVE THREAT | Vendor-confirmed exploitation

 

Check Point’s investigation links PHP2Shell initial access on Mobile Access portals to exploitation of DynamicCLI, allowing the portal service account to execute as root. Observed activity included web shells, the TSGA backdoor, proxying and a destructive phase that reached VMware ESXi in some incidents.

 

Attack path: Internet-facing portal compromise → local service-account access → DynamicCLI privilege escalation → root on the gateway → persistence and downstream pivot.

 

Why it matters: Local privilege escalation on a security gateway is not a low-scope event when remote initial access exists. It converts a web-facing weakness into authority over a network enforcement point.

 

Remediation and validation: Install the applicable Jumbo Hotfix, confirm the exact gateway take, and hunt the full chain rather than the CVE in isolation. Review portal logs, DynamicCLI activity, web roots, startup mechanisms, outbound tunnels and connections from the gateway to virtualization or management networks.

 

6. Perforce Puppet Enterprise Report-Deserialization RCE — CVE-2026-115731

NEW THIS WEEK | CISA-confirmed exploitation

 

Affected Puppet Server code accepts a crafted rich-data report from a client with any CA-signed agent certificate and can execute code as the service user. In environments with permissive autosigning, a remote attacker may be able to obtain the certificate needed to cross that boundary.

 

Attack path: Valid or improperly autosigned agent certificate → crafted report PUT request → deserialization/code execution on Puppet Server → access to automation trust and configuration data.

 

Why it matters: Configuration-management infrastructure has authority over many endpoints. Server compromise may expose code, secrets and relationships that expand the incident well beyond one host.

 

Remediation and validation: Upgrade Puppet Enterprise to 2023.8.13, 2025.12.1, 2026.0.1 or later applicable fixes and restart Puppet Server. Audit autosign policy, certificate requests and revocations; examine unusual report payloads and process execution; and confirm that downstream credentials stored or reachable from the server remain trustworthy.

 

7. Apache Axis Unsafe Reflection RCE — CVE-2026-90436

NEW THIS WEEK | CISA-confirmed exploitation

 

Apache Axis 1.x parses a crafted SOAP fault and loads an attacker-named class before per-service handler chains run. If a suitable class exists on the application classpath, class initialization or instantiation can produce remote code execution. Axis 1.x is end-of-life, and Apache does not expect a supported 1.x release to resolve the issue.

 

Attack path: Crafted SOAP fault → attacker-selected class load / string-constructor instantiation → classpath-dependent code execution before service authentication.

 

Why it matters: The vulnerable library may be nested inside a WAR, EAR or vendor application. Traditional asset inventory can miss it, and a compensating authentication handler may run too late.

 

Remediation and validation: Search build manifests, SBOMs and deployed Java archives for Axis 1.x. Migrate to a supported SOAP engine. Where migration cannot be immediate, use a controlled build containing the upstream patch, restrict endpoint reachability and validate that affected classes cannot be invoked.

 

8. Synology DSM/BSM System Plug-In RCE — CVE-2024-10441

NEW THIS WEEK | CISA-confirmed exploitation

 

Synology’s advisory describes remote code execution in the system plug-in daemon and supplies fixed releases across DSM and BeeStation OS. CISA’s KEV addition on November 18 establishes exploitation; the public record still does not provide enough vector detail to characterize the exact request path responsibly.

 

Attack path: Remote access to an affected DSM/BSM service → unspecified input path in the plug-in daemon → arbitrary code execution.

 

Why it matters: Storage systems hold production data, backups and credentials. Compromise can affect confidentiality and recovery readiness simultaneously.

 

Remediation and validation: Upgrade to the fixed Synology release listed for the deployed branch, reduce internet and management-plane exposure, review administrative accounts and packages, and inspect scheduled tasks, file access and outbound traffic. Do not substitute speculative detection logic for the vendor fix when the public exploit vector is incomplete.

 

9. Vanderbilt REDCap Survey-Context Code Injection — CVE-2026-90817

NEW THIS WEEK | CISA-confirmed exploitation

 

The vulnerability abuses survey passthrough routing and data-import handling. An unauthenticated attacker needs a valid public survey hash, then can craft requests that reach an unintended controller route and supply a malicious path or stream parameter for server-side code execution.

 

Attack path: Knowledge of a public survey hash → manipulated survey-context request → unintended data-import controller → code execution.

 

Why it matters: Public surveys are intentionally reachable and often handle sensitive health or research data. A survey identifier becomes part of an authority path when server routing fails to preserve the trust boundary.

 

Remediation and validation: Use the REDCap institutional support process to determine the fixed release and apply it promptly. Inventory externally reachable surveys, retire or rotate unnecessary links, monitor controller and import activity, and involve application and data owners in the exposure decision.

 

10. SAP NetWeaver Enterprise Portal JNDI Injection — CVE-2025-42884

NEW THIS WEEK | CISA-confirmed exploitation

 

The flaw allows unauthenticated manipulation of JNDI lookup properties or URLs. SAP’s CVSS assessment focuses on disclosure or modification of server information; Onapsis explains that remote code execution is possible only when a suitable gadget chain is available and not blocked by deserialization filters. SAP Note 3660969 has provided the patch since November 2025.

 

Attack path: Reachable Enterprise Portal component → injected JNDI properties / external lookup → server-information impact and, in qualifying Java environments, potential deserialization-based code execution.

 

Why it matters: Base severity does not describe the business importance of a portal connected to enterprise SAP services. A patch that has existed for a year can remain absent from inherited or hard-to-change systems.

 

Remediation and validation: Apply SAP Note 3660969 to EP-RUNTIME and EP-BASIS, implement the measures in Note 3660659, restrict portal and JNDI egress exposure, and monitor unexpected lookup destinations. Validate the component level in the running landscape rather than relying on a central patch record.

 

What Changed This Week

 

CISA added nine vulnerabilities to KEV during the reporting window: Perforce Puppet Enterprise on November 17; Apache Axis, REDCap, LiteLLM and Synology on November 18; and Cisco Expressway, SAP NetWeaver, VMware vCenter and Apache Roller on November 20.

 

Eight appear in this Top 10. LiteLLM CVE-2026-40217 remains operationally relevant but did not outrank the continuing Barracuda or Check Point paths. Its default configuration requires a proxy-admin credential to reach the vulnerable custom-code guardrail endpoint. Organizations using affected LiteLLM versions should upgrade to 1.83.11 or later, or block POST /guardrails/test_custom_code at the reverse proxy until they can.

 

The more important pattern is authority concentration. vCenter controls virtual infrastructure and identity. Puppet distributes configuration. Expressway and security gateways bridge trust boundaries. Storage and research platforms concentrate sensitive data. The likely consequence is determined by what the system can authorize after compromise, not only the initial technical primitive.

 

Executive Priorities

 

  1. Interrupt the highest-authority paths

    Treat vCenter, Expressway, Puppet and security gateways as control-plane systems. Patch them ahead of lower-authority assets and use temporary network restrictions when the permanent change cannot happen immediately.

  2. Verify compromise, not just vulnerability state

    For vCenter, Check Point and Barracuda, exploitation may leave administrators, keys, web shells or configuration changes that survive the patch. Define separate workstreams for remediation and incident validation.

  3. Reduce management exposure

    Administrative interfaces, database services, XML-RPC endpoints and SOAP services should not be broadly reachable. Use dedicated management segments, allowlists and authenticated access paths, then test those controls from the attacker’s side of the boundary.

  4. Find embedded and inherited software

    Axis 1.x, old SAP portal components and long-running Synology appliances can escape modern inventory. Use software composition data, network observation and owner interviews together. An “unknown owner” is a risk decision waiting to be made.

  5. Demand evidence of closure

    For every priority remediation, require the running version, service restart status, endpoint retest, management-exposure check, telemetry review and named owner. Closure should demonstrate that the attack path no longer works and that persistence was considered.

 

Leadership POV

 

Vulnerability leadership is an authority-allocation discipline. The same technical flaw means something different on an isolated utility than on a system that can create SSO administrators, configure thousands of endpoints, enforce the perimeter or hold the recovery copy of business data.

Leaders should make three decisions explicit: which attack path is interrupted first, who owns the system and adjacent controls, and what evidence will prove that enterprise exposure declined. Strong foundations—accurate inventory, constrained administration, supported software, trustworthy identity and observable change—make those decisions faster and less dependent on crisis response.

 

Patch the control plane. Contain the edge. Verify authority.

 

Verification and Classification Method

 

This analysis uses only public information available by Sunday, November 22, 2026 at 12:05 AM America/Chicago (CST). Ranking considers confirmed exploitation, exposure, prerequisites, authority gained, blast radius, prevalence, exploit maturity, remediation complexity and business consequence. It does not rank by CVSS alone.

 

NEW THIS WEEK means newly qualifying for this edition based on evidence published during November 15–21 and comparison with the prior November 8–14 edition. It does not assert that the vulnerability or patch first existed this week. Barracuda CVE-2026-116606 and Check Point CVE-2026-123039 are CONTINUING ACTIVE THREATS.

 

ZERO-DAY is reserved for VMware CVE-2026-123890 because Mandiant’s evidence dates exploitation to early September and Broadcom’s fix to November 19. Apache Roller is actively exploited, but the public record does not state when exploitation began relative to the September 24 release of Roller 6.1.6; no zero-day label is applied. Public exploitation detail is limited for several KEV entries, especially Synology and REDCap. Those gaps are stated rather than filled with inference.

 

Sources & References

  1.  CISA Known Exploited Vulnerabilities Catalog: https://www.cisa.gov/known-exploited-vulnerabilities-catalog

  2. Google Threat Intelligence Group / Mandiant — ARMORY CASTLE Exploits vSphere: https://cloud.google.com/blog/topics/threat-intelligence/armory-castle-exploits-vsphere

  3. Broadcom VMSA-2026-0009.1 — VMware vCenter updates: https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/VMSA-2026-0009--VMware-vCenter--Tools--and-SDDC-Manager-updates-address-multiple-vulnerabilities--CVE-2026-123890--CVE-2026-123891--CVE-2026-123892--CVE-2026-123893--CVE-2026-123894-/39495

  4. Cisco Security Advisory — Expressway Series Remote Code Execution: https://www.cisco.com/c/en/us/support/docs/csa/cisco-sa-vcs-static-cred-hr7EWZ8t.html

  5. Barracuda — Pre-Authentication Authorization Bypass and Command Injection: https://trust.barracuda.com/security/pre-authentication-authorization-bypass-command-injection

  6. Apache Roller advisory — CVE-2026-129448: https://lists.apache.org/thread/lnz4s26tv6hdo03wngtdbpfj4hg4gtds

  7. Check Point Research — PHP2Shell Exploitation on Check Point Gateway: https://research.checkpoint.com/2026/php2shell-exploitation-on-check-point-gateway/

  8. Perforce Puppet Enterprise — CVE-2026-115731: https://portal.perforce.com/s/cve/a91Qi000003IaHpIAK/report-deserialization-code-execution

  9. Apache Axis advisory — CVE-2026-90436: https://lists.apache.org/thread/8slf0j9gz35z56t0glt5h0ncwlbn4bdz

  10. Synology SA-24:20 — CVE-2024-10441: https://www.synology.com/en-global/security/advisory/Synology_SA_24_20

  11. SAP Security Patch Day Bulletins — CVE-2025-42884 / Note 3660969: https://support.sap.com/en/my-support/knowledge-base/security-notes-news/bulletin-2025.html

  12. Onapsis — SAP NetWeaver unauthenticated JNDI injection: https://onapsis.com/security-advisories/sap-netweaver-java-unauthenticated-jndi-injection-in-sap-netweaver-enterprise-portal/

Debra A. Anderson, MBA

Debra A. Anderson is an enterprise cybersecurity and exposure-management leader with more than 40 years of business experience, 30 years in information technology, and 18 years in cybersecurity.

Her work focuses on connecting technology to business outcomes through vulnerability and exposure management, operational validation, cyber risk reduction, and AI governance. She brings practical experience managing complex enterprise environments, improving remediation processes, and translating technical findings into decisions business leaders can act on.

Through her IN THE WILD series, Debra examines actively exploited vulnerabilities, operational exposure, and the evidence organizations need to determine whether risk has actually been reduced.

Her guiding principle is simple: effective cybersecurity is measured by verified outcomes, not just completed activities.

https://www.linkedin.com/in/debraanderson-infosec/
Next
Next

HAWK AI THREAT BRIEF | TOP 10 AI DEVELOPMENTS SECURITY TEAMS SHOULD BE WATCHING