CVE-2026-116506: command injection

CISO Take

A command injection flaw in the apmib_set function of firewall.so on the TOTOLINK A950RG router (firmware 4.1.2cu.5268_B20221104) lets a remote, low-privileged user run arbitrary commands on the device. NVD rates it critical (CVSS 9.9, scope changed), and the researcher has publicly disclosed the technique. This is consumer or SMB networking gear, not an AI/ML product, and the 'ai_tool' category looks like a false-positive classification. Its relevance to AI environments is indirect: a compromised router can sit in the traffic path of AI workloads, so it can expose model APIs, credentials, or data flows. The CVE is not in CISA KEV, no EPSS score is available yet, and our feed shows no Nuclei template or tracked exploit, although the vendor-neutral disclosure and VulDB entries mean exploitation is likely to be easy. Check whether any A950RG units are deployed, keep their management interfaces off the internet, and monitor the vendor site for a firmware fix, since none is referenced.

Sources:NVDVulDBATLAS

What is the risk?

Technical severity is high (CVSS 9.9: network, low complexity, low privileges, no user interaction, scope changed, high C/I/A). Exploitation likelihood is plausibly high because the exploit is publicly disclosed and the flaw is a simple command injection, but there is no EPSS score, KEV listing, SSVC decision, or scanner template to confirm in-the-wild use. Exposure to AI/ML systems is indirect and depends on whether the router fronts AI infrastructure such as home labs, small offices, or edge inference sites. Confidence that this is not an AI-specific vulnerability: HIGH. Confidence in the exploitability assessment: MEDIUM.

How does the attack unfold?

Initial Access

Attacker reaches the router's management interface and authenticates with a low-privileged account.

AML.T0049

Exploitation

A crafted request injects shell commands through apmib_set in firewall.so, causing OS command execution.

AML.T0050

Persistence and Pivot

Attacker plants a backdoor or reverse shell and uses the router to reach internal hosts, including AI services.

Impact

Traffic interception, credential theft, or loss of network integrity and availability affecting connected AI workloads.

Initial Access

Attacker reaches the router's management interface and authenticates with a low-privileged account.

AML.T0049

Exploitation

A crafted request injects shell commands through apmib_set in firewall.so, causing OS command execution.

AML.T0050

Persistence and Pivot

Attacker plants a backdoor or reverse shell and uses the router to reach internal hosts, including AI services.

Impact

Traffic interception, credential theft, or loss of network integrity and availability affecting connected AI workloads.

How severe is it?

CVSS 3.1

9.9 / 10

EPSS

1.9%

chance of exploitation in 30 days

Higher than 79% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Trivial

What is the attack surface?

AVNetwork

ACLow

PRLow

UINone

SChanged

CHigh

IHigh

AHigh

What should I do?

1 step

1) Inventory networks for TOTOLINK A950RG devices, especially firmware 4.1.2cu.5268_B20221104, using asset discovery and network scans. 2) Check totolink.net for a patched firmware release and apply it. None is referenced in the CVE data, so verify rather than assume. 3) Disable remote (WAN-side) administration and restrict the web management interface to a dedicated management VLAN or trusted IPs. 4) Change default and weak credentials, since exploitation requires a low-privileged authenticated account. 5) If no patch is available, replace or isolate the device and keep it away from segments hosting AI workloads or sensitive data. 6) Monitor for unexpected outbound connections, new processes or listeners on the device, configuration changes, and DNS changes. Review the referenced technical write-up (the gist and VulDB entries) to build detections for the request pattern.

How is it classified?

Code ExecutionFrameworkAML.T0049- Exploit Public-Facing Application AML.T0050- Command and Scripting Interpreter

Which compliance frameworks are affected?

This CVE is relevant to:

ISO 42001

A.10.3-Suppliers

NIST AI RMF

MANAGE 2.3-Responses to previously unknown risks and supply chain/third-party components

Frequently Asked Questions

What is CVE-2026-116506?

A command injection flaw in the apmib_set function of firewall.so on the TOTOLINK A950RG router (firmware 4.1.2cu.5268_B20221104) lets a remote, low-privileged user run arbitrary commands on the device. NVD rates it critical (CVSS 9.9, scope changed), and the researcher has publicly disclosed the technique. This is consumer or SMB networking gear, not an AI/ML product, and the 'ai_tool' category looks like a false-positive classification. Its relevance to AI environments is indirect: a compromised router can sit in the traffic path of AI workloads, so it can expose model APIs, credentials, or data flows. The CVE is not in CISA KEV, no EPSS score is available yet, and our feed shows no Nuclei template or tracked exploit, although the vendor-neutral disclosure and VulDB entries mean exploitation is likely to be easy. Check whether any A950RG units are deployed, keep their management interfaces off the internet, and monitor the vendor site for a firmware fix, since none is referenced.

Is CVE-2026-116506 actively exploited?

No confirmed active exploitation of CVE-2026-116506 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-116506?

1) Inventory networks for TOTOLINK A950RG devices, especially firmware 4.1.2cu.5268_B20221104, using asset discovery and network scans. 2) Check totolink.net for a patched firmware release and apply it. None is referenced in the CVE data, so verify rather than assume. 3) Disable remote (WAN-side) administration and restrict the web management interface to a dedicated management VLAN or trusted IPs. 4) Change default and weak credentials, since exploitation requires a low-privileged authenticated account. 5) If no patch is available, replace or isolate the device and keep it away from segments hosting AI workloads or sensitive data. 6) Monitor for unexpected outbound connections, new processes or listeners on the device, configuration changes, and DNS changes. Review the referenced technical write-up (the gist and VulDB entries) to build detections for the request pattern.

What systems are affected by CVE-2026-116506?

This vulnerability affects the following AI/ML architecture patterns: Network edge / SOHO routers, Self-hosted AI workloads behind vulnerable routers (indirect), Remote-access paths to AI development environments (indirect).

What is the CVSS score for CVE-2026-116506?

CVE-2026-116506 has a CVSS v3.1 base score of 9.9 (CRITICAL). The EPSS exploitation probability is 1.88%.

What is the AI security impact?

Affected AI Architectures

Network edge / SOHO routersSelf-hosted AI workloads behind vulnerable routers (indirect)Remote-access paths to AI development environments (indirect)

MITRE ATLAS Techniques

AML.T0049Exploit Public-Facing Application

AML.T0050Command and Scripting Interpreter

Compliance Controls Affected

ISO 42001: A.10.3

NIST AI RMF: MANAGE 2.3

What are the technical details?

Original Advisory

A vulnerability was determined in TOTOLINK A950RG 4.1.2cu.5268_B20221104. Affected is the function apmib_set of the file firewall.so. This manipulation causes command injection. The attack may be initiated remotely. The exploit has been publicly disclosed and may be utilized.

Exploitation Scenario

An attacker identifies an internet-reachable or LAN-reachable A950RG, authenticates with a low-privileged or default account, and sends a crafted request that reaches apmib_set in firewall.so, injecting shell metacharacters into a firewall-related configuration field. This yields command execution on the router. The attacker then installs a reverse shell or persistent backdoor, alters DNS or routing, and intercepts traffic to a nearby self-hosted AI service (such as an exposed inference API or notebook server) to harvest API keys or tokens, or uses the router as a pivot or botnet node.

Weaknesses (CWE)

CWE-74Improper Neutralization of Special Elements in Output Used by a Downstream Component ('Injection')Primary CWE-77Improper Neutralization of Special Elements used in a Command ('Command Injection')Primary CWE-74Improper Neutralization of Special Elements in Output Used by a Downstream Component ('Injection')CWE-77Improper Neutralization of Special Elements used in a Command ('Command Injection')

CWE-74 — Improper Neutralization of Special Elements in Output Used by a Downstream Component ('Injection'): The product constructs all or part of a command, data structure, or record using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify how it is parsed or interpreted when it is sent to a downstream component.

[Requirements] Programming languages and supporting technologies might be chosen which are not subject to these issues.
[Implementation] Utilize an appropriate mix of allowlist and denylist parsing to filter control-plane syntax from all input.

Source: MITRE CWE corpus.