Control console without access protection

NASA Software AIT-GUI Allowed Spacecraft Commands Without any Login

NASA
Facebook
X
LinkedIn
Reddit
WhatsApp
Source: SNEHIT PHOTO/Shutterstock.com

Security researchers have disclosed a vulnerability chain in AIT-GUI, the browser-based control console for NASA/JPL’s open-source AMMOS Instrument Toolkit.

Tracked as GHSA-p9r8-2q67-fp86 and rated with a CVSS score of 9.4, the vulnerability chain affects AIT-GUI version 2.5.1 and earlier; it was fixed in version 2.5.2 on August 12. According to the security advisory published on August 13, the AIT-GUI web server ignored the host address that was actually configured and instead bound itself statically to the address 0.0.0.0 on port 8080, making it reachable on every network interface. All state-changing routes of the application were secured neither by authentication nor by authorization or protection against cross-site request forgery. Cycode summarized the impact as follows:

Ad

“The impact of an unauthenticated request is measured in executed instrument commands, not defaced web pages.”

Cycode

Ad

Specifically, arbitrary instrument and spacecraft commands could be issued via the affected routes, server-side scripts including those outside the intended directory could be executed, and command sequences could be launched, also with the ability to pass files outside the intended directory to a subprocess. Since the affected routes accepted form data that browsers treat as so-called simple requests without prior preflight checks, visiting a crafted website by an operator with access to the console was sufficient to trigger a successful attack, according to Cycode, even on an installation that was actually only accessible locally or behind a firewall.

Two different advisories with conflicting information

Version 2.5.2 now binds the server to the configured, by default local address and compares the Origin or Referer header with its own host during state-modifying requests, preventing CSRF attacks via the browser. However, an independent review of the source code by The Hacker News revealed that the central route function continues to issue a valid session without access checks, and the command route accepts any request with such a session cookie, meaning the actual authentication vulnerability persists.

In parallel, a second entry published independently by VulnCheck and the National Vulnerability Database back on July 29, under the ID CVE-2026-60112 and with a CVSS score of 9.3, documents the same fundamental authentication flaw in the same file. Both advisories contradict each other regarding the actual bug fix: While VulnCheck and the NVD entry cite a commit made back in July as the fix, the Cycode advisory continues to list version 2.5.1 as vulnerable and only 2.5.2 as corrected.

According to The Hacker News’s own verification, the unprotected session issuance was indeed still present in both version 2.5.1 and version 2.5.2. Additionally, a check against the Python Package Index (PyPI) showed that only version 2.4.1 from July 2023 has been published there so far, meaning the newer versions 2.5.0 through 2.5.2 mentioned in the security advisories are not available via the standard pip installation method at all.

NASA: Fix developed with AI assistance

Cycode security researcher Yuval Elbar authored the actual commit mitigating the vulnerability back on July 10, including a co-authorship credit to Anthropic’s Claude Opus 4.8 model as well as 18 additional regression tests specifically written for the flaw. Cycode itself framed the discovery as exemplary of future security research, in which human researchers increasingly work alongside AI-assisted code analysis.

Neither advisory names a specific space mission using the software, nor is there any evidence of actual exploitation of the vulnerability based on current knowledge. In addition, the GitHub Advisory Database contains several other, as yet not fully reviewed entries regarding other NASA ground software, as well as CVE-2024-35058, an unpatched critical vulnerability originally published in 2024 in the toolkit’s underlying AIT-Core libraries.

(Editorial Team)

Ad

Weitere Artikel