CVE-2026-110455

Executive Summary

CVE-2026-110455 is an Improper Input Validation vulnerability (CWE-20) in Ubiquiti's UniFi Protect AI Key that can result in command injection on the affected device. A malicious actor with network access, given specific preconditions, can exploit this flaw to execute arbitrary commands on the device at the underlying operating system level. With a CVSS base score of 8.1, the vulnerability poses a high confidentiality, integrity, and availability impact, though exploitation complexity is rated as High (AC:H), suggesting that attack conditions must be carefully met.

Technical Analysis

The vulnerability is classified as CWE-20 (Improper Input Validation) and manifests as a command injection issue within the UniFi Protect AI Key. Command injection vulnerabilities arise when an application passes unsanitized user-controllable input to a system shell or command execution function. In this case, the NVD description explicitly states that exploitation leads to "Command Injection on the device," meaning the attacker can execute arbitrary operating system commands in the context of the vulnerable service.

The attack vector is network-based (AV:N), requiring no authentication (PR:N) and no user interaction (UI:N). The lack of required privileges indicates that the vulnerable input-handling functionality is reachable pre-authentication or through a low-privilege network position. However, the High attack complexity (AC:H) rating implies that exploitation is not trivially repeatable — it likely depends on specific conditions such as race conditions, particular device states, or specially crafted request sequencing that may not be universally reproducible.

Given that this is an AI Key device within the UniFi Protect ecosystem, the attack surface is likely a network-facing service (HTTP/HTTPS management interface, API, or inter-device communication protocol) that accepts input which is subsequently incorporated into a system-level command without adequate validation. The exact endpoint, parameter names, and internal function names are not disclosed in the available intelligence; no CVE description details, CNA title, or technical write-up has been provided that specifies the vulnerable code path. Therefore, the precise injection point cannot be confirmed from the data at hand.

Because the CVSS vector yields Scope:Unchanged (S:U), the impact is confined to the vulnerable device's security context. However, with High impacts across all three CIA pillars, successful exploitation would grant the attacker full control over the AI Key device, including the ability to exfiltrate data, alter configurations, or disrupt availability.

The affected product is Ubiquiti UniFi Protect AI Key. Affected version ranges, OSV package identifiers, and vendor advisory references are not available in the provided intelligence. The vendor is not listed as having published a fix, and no CISA KEV entry exists, indicating that this vulnerability is not currently known to be exploited in the wild at scale. EPSS scoring has not been assigned by FIRST.org at the time of this report.

Exploitation Scenarios

The absence of a public proof-of-concept, Nuclei template, or exploit code means the following attack chain is inferred from the vulnerability class (improper input validation leading to command injection) and the CVSS vector. No specific function names, API endpoints, or parameter names are provided in the source data and none are fabricated here.

1. Reconnaissance / Target Identification: The attacker identifies a UniFi Protect AI Key device exposed on the network. This could be achieved through network scanning for the device's management or communication interfaces, Shodan queries for internet-exposed instances, or lateral movement within a compromised network segment. The device could be a standalone AI Key appliance or integrated into a broader UniFi Protect deployment.

2. Triggering the Vulnerability Class: The attacker crafts input designed to break out of the expected data context and inject operating system commands. The specific input vector — whether an HTTP request parameter, a header field, a serialized payload, or an inter-process message — is not disclosed in the available intelligence. Because the flaw is CWE-20, the application fails to properly validate, sanitize, or escape the input before passing it to a command execution context (e.g., a shell call, system() invocation, or equivalent mechanism). The attacker would need to satisfy the High attack complexity condition, which may involve timing, specific device state, or particular request formatting.

3. Payload Execution: At the vulnerability-class level, the injected input is interpreted by the device's command interpreter, resulting in the execution of attacker-supplied commands. This typically allows the attacker to run shell commands, spawn reverse shells, download and execute secondary payloads, or manipulate the device's file system. Because the vulnerability allows command injection, the attacker's commands execute with the privileges of the vulnerable service process — potentially root or a highly privileged account depending on the device's architecture.

4. Post-Exploitation Impact: Successful command injection on the UniFi Protect AI Key grants the attacker a foothold on the device. From there, the attacker could:

   • Exfiltrate sensitive video metadata, encryption keys, or configuration data processed by the AI Key.
   • Pivot to other devices on the network, potentially including UniFi Protect cameras, NVRs, or the broader UniFi management plane.
   • Disable or tamper with AI-driven analytics functions (e.g., person/vehicle detection) to facilitate physical security bypass.
   • Establish persistence for long-term access.
   • Use the device as a relay for further network attacks.

The High attack complexity suggests that not every attempt will succeed, and the attacker may need to iterate or possess specific knowledge of the device's internal state. No public exploit exists, so the attack chain remains theoretical pending further technical disclosure.

Impact Assessment

• Confidentiality (C:H): Successful command injection allows full read access to data on the device, including potentially sensitive AI Key material, Protect configuration, and network credentials.
• Integrity (I:H): The attacker can modify files, alter configurations, and tamper with the device's operation.
• Availability (A:H): The attacker can disrupt or disable the device, impacting the UniFi Protect surveillance and analytics capability.
• Scope (S:U): The impact is confined to the vulnerable device's security authority, though lateral movement remains a post-exploitation risk.
• Attack Vector (AV:N): Exploitable from the network, including potentially from the internet if the device is exposed.
• Attack Complexity (AC:H): Exploitation requires specific conditions that are not under the attacker's full control, reducing the likelihood of automated mass exploitation.
• Privileges Required (PR:N): No authentication is needed, meaning the attacker does not require valid credentials.
• User Interaction (UI:N): No user action is required for exploitation.

The blast radius includes any organization using the affected UniFi Protect AI Key. Given the device's role in physical security and AI-driven video analytics, a compromise could have significant operational and security implications, especially in environments where the Protect system is used for critical surveillance or access control decisions. The CVSS score of 8.1 reflects a high-severity vulnerability, but the High attack complexity and lack of public exploit code may limit immediate widespread exploitation. However, a motivated attacker with knowledge of the device internals could develop a working exploit.

Detection Guidance

Sigma Rule

title: Potential Command Injection Attempt Against UniFi Protect AI Key (CVE-2026-110455)
id: 8f4c2a1e-9b3d-4e7f-a1c2-6d5e8f9a0b1c
status: experimental
description: Detects network requests to UniFi Protect AI Key devices containing common command injection patterns. This rule is speculative as no specific endpoint or parameter is known.
references:
  - https://nvd.nist.gov/vuln/detail/CVE-2026-110455
author: Threat Intelligence
date: 2026-06-01
tags:
  - attack.t1190
  - attack.t1059
logsource:
  category: webserver
  product: unifi
detection:
  selection:
    cs-uri-query|contains:
      - ';'
      - '|'
      - '`'
      - '$('
      - '&&'
      - '||'
      - '../'
      - '..\\'
  condition: selection
falsepositives:
  - Legitimate use of special characters in query parameters
level: high

YARA Rule

rule CVE_2026_110455_CommandInjection_UniFi_Protect_AI_Key
{
    meta:
        description = "Detects potential command injection payloads targeting UniFi Protect AI Key (CVE-2026-110455)"
        author = "Threat Intelligence"
        date = "2026-06-01"
        reference = "https://nvd.nist.gov/vuln/detail/CVE-2026-110455"
        severity = "high"

    strings:
        $s1 = "UniFi Protect" ascii wide nocase
        $s2 = "AI Key" ascii wide nocase
        $cmd1 = "cmd=" ascii wide nocase
        $cmd2 = "exec=" ascii wide nocase
        $cmd3 = "/bin/sh" ascii wide
        $cmd4 = "/bin/bash" ascii wide
        $cmd5 = "system(" ascii wide
        $cmd6 = "popen(" ascii wide

    condition:
        any of ($s*) and any of ($cmd*)
}

Suricata IDS Rule

alert http any any -> any any (msg:"CVE-2026-110455 Potential Command Injection Attempt Against UniFi Protect AI Key"; flow:to_server,established; content:"UniFi"; http_header; nocase; content:"AI Key"; http_header; nocase; pcre:"/(\%3B|;|\||`|\$\(|\&\&|\|\|)/Ui"; threshold:type limit, track by_src, count 1, seconds 60; classtype:attempted-admin; sid:2026110455; rev:1;)

Elastic Detection

network where
  process.name != null and
  process.command_line != null and
  (
    process.command_line : ("*/bin/sh*", "*/bin/bash*", "*system(*", "*popen(*", "*cmd=*", "*exec=*")
  ) and
  (
    process.parent.name : ("*unifi*", "*protect*", "*aikey*")
  )

Splunk SPL Query

index=web OR index=proxy
| search (uri_path="*" OR uri_query="*")
| where match(uri_query, "(?i)(;|\||`|\$\(|\&\&|\|\|)")
| stats count values(uri_query) as queries by src_ip, dest_ip, uri_path
| where count > 0
| table src_ip, dest_ip, uri_path, count, queries

Wazuh OSSEC Rule

<group name="unifi,protect,cve-2026-110455,">
  <rule id="11045500" level="12">
    <decoded_as>json</decoded_as>
    <field name="url">\.*(;|\||`|\$\(|\&\&|\|\|)\.*</field>
    <description>CVE-2026-110455: Possible command injection attempt against UniFi Protect AI Key</description>
    <mitre>
      <id>T1190</id>
      <id>T1059</id>
    </mitre>
  </rule>
</group>

Mitigation & Patch Guidance

Patch confidence: 0/4 (UNPATCHED — no patch evidence from any tracked source).

1. Immediate Mitigations:

   • Restrict network access to the UniFi Protect AI Key device. Place it behind a firewall or VLAN segmentation, allowing only trusted management hosts to reach its services.
   • Disable or block any unnecessary network services on the device, particularly those reachable from untrusted network segments.
   • Monitor for anomalous outbound connections or command execution artifacts on the device (e.g., unexpected processes, cron jobs, or files).
   • If the device is internet-exposed, immediately remove it from public reachability.
   • Apply strict input validation at network perimeter devices (WAF/IPS) if traffic passes through them, focusing on command injection patterns.

2. Patch/Upgrade Steps:

   • No vendor patch is currently available per tracked sources. Monitor Ubiquiti's official security advisory page and UniFi Protect release notes for updates.
   • Once a patch is released, apply it immediately following the vendor's instructions. Upgrade the AI Key firmware to the fixed version when available.

3. Long-Term Hardening:

   • Enforce the principle of least privilege on the device where possible.
   • Implement network segmentation for all IoT/physical security devices.
   • Conduct regular vulnerability scanning and firmware audits of UniFi Protect components.
   • Deploy the detection rules provided in this report and tune them to reduce false positives.
   • Consider replacing the device with an alternative if no patch is forthcoming and the risk is deemed unacceptable.

References

1. NVD – CVE-2026-110455: https://nvd.nist.gov/vuln/detail/CVE-2026-110455
2. Ubiquiti Security Advisory Portal: https://www.ui.com/security-advisories/
3. MITRE ATT&CK T1190 – Exploit Public-Facing Application: https://attack.mitre.org/techniques/T1190/
4. MITRE ATT&CK T1059 – Command and Scripting Interpreter: https://attack.mitre.org/techniques/T1059/
5. CWE-20 – Improper Input Validation: https://cwe.mitre.org/data/definitions/20.html

No known threat actor campaigns or exploit-in-the-wild evidence have been reported for this CVE. Absence of public PoC and KEV listing indicates a lower immediate exploitation risk, but the high severity warrants proactive defense.