Description

Looking Glass is a modern, stateless network-diagnostic platform — a single self-contained Go binary that fronts a fleet of routers over SSH and exposes ping / traceroute / BGP lookups through a gRPC (ConnectRPC) API, an embedded SvelteKit web UI, and a lg-cli client. Prior to version 1.3.5, there is an OS Command Injection vulnerability resulting from an unanchored regular expression in the input validation layer. This issue has been patched in version 1.3.5.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "9.8",
  "pubDate": "2026-09-02T16:17:18.127Z",
  "pubdate": "2026-09-02T16:17:18.127Z",
  "executiveSummary": "Looking Glass versions prior to 1.3.5 contain an OS Command Injection vulnerability arising from improper input validation within the system's request handling layer.\nThe vulnerability originates from the use of an unanchored regular expression intended to sanitize user-supplied input for network diagnostic operations such as ping and traceroute.\nAn unauthenticated remote attacker can exploit this flaw by supplying specially crafted diagnostic parameters that bypass validation filters, facilitating the injection of arbitrary shell commands.\nSuccessful exploitation results in Remote Code Execution (RCE) with the privileges of the Looking Glass service process, potentially leading to full system compromise, unauthorized access to internal network infrastructure through SSH, and exfiltration of sensitive diagnostic data.\nThis vulnerability poses a critical risk to deployments where the platform is exposed to untrusted network segments, as it allows attackers to leverage the tool's intended functionality to execute unauthorized operating system commands.",
  "technicalDetails": "The vulnerability resides within the input validation logic of the Looking Glass platform, which is responsible for sanitizing parameters passed to network diagnostic utilities like ping, traceroute, and BGP lookups.\nThe root cause is identified as an unanchored regular expression used to validate user input. In regular expression matching, an unanchored pattern does not enforce constraints across the entire input string. Specifically, if a regex pattern does not start with a start-of-line anchor (^) or end with an end-of-line anchor ($), an attacker can inject malicious shell metacharacters—such as semicolons (;), pipes (|), or backticks (`)—before or after the expected valid input pattern.\nWhen the Looking Glass service processes a diagnostic request via the ConnectRPC API or the embedded SvelteKit web UI, the improperly sanitized string is passed to the underlying operating system to invoke external network utilities via SSH or local shell execution.\nThe attack flow follows a predictable pattern: 1) The attacker constructs a crafted HTTP request containing diagnostic parameters that include shell injection payloads. 2) The application's input validation layer evaluates the payload against the flawed unanchored regex. 3) Because the regex fails to validate the entirety of the input, the malicious payload is permitted to reach the backend command execution function. 4) The backend process interprets the injected shell metacharacters, executing the attacker-supplied commands concurrently with or in place of the intended diagnostic command.\nThis exploitation allows for the execution of arbitrary commands with the binary's runtime privileges. Given that Looking Glass operates by fronting a fleet of routers over SSH, an attacker may leverage this access to pivot into the broader network infrastructure, intercept diagnostic output, or maintain persistence on the host machine.\nThe flaw affects all deployments of the Looking Glass Go binary versioned prior to 1.3.5. As the platform exposes these functions via gRPC and web interfaces, the attack surface includes any entity capable of reaching the platform's API endpoints, regardless of authentication state, unless strictly isolated at the network boundary."
}