The flaw exists because several input fields are modeled as ordinary text even though they are later written directly into line-oriented configuration files. When a user places CR or LF characters in a value expected to contain one line, the value can terminate the intended configuration line and create additional attacker-controlled directives or policy lines in a privileged service configuration.
In the confirmed Dnsmasq path, dhcp_boot.filename, dhcp_boot.servername, dhcp_boot.address, and dhcp_options.value are rendered into /usr/local/etc/dnsmasq.conf. The relevant access includes the page-services-dnsforwarder ACL and APIs under api/dnsmasq/*. The template writes these values directly after Dnsmasq directives, while the fields do not receive sufficient single-line validation before storage.
The attack requires an authenticated Web UI or API account with delegated permission to configure an affected component, plus the ability to apply or restart the service so that the generated configuration is loaded. The same pattern is documented for OpenVPN, DHCPv6, Unbound, IPsec, NTP, PPP or MPD, Monit, firewall, sudo, WebGUI, and System General configuration, but the exact result depends on each service's syntax and behavior. Arbitrary shell execution was not confirmed for the Dnsmasq path; the Monit path demonstrated command execution with root privileges.