Critical vulnerability in NASA open-source software allowed sending commands to spacecraft without a password

Cyber analysts at Cycode discovered a critical chain of vulnerabilities in NASA's open-source software — the AMMOS Instrument Toolkit (AIT) — that allowed unauthorized users to send arbitrary commands to spacecraft and execute arbitrary scripts on the server. This is reported by RBC-Ukraine, citing a technical report by Cycode. The AIT ground system is used to build telemetry processing systems and transmit commands to spacecraft, making any gaps in its security potentially catastrophic for active missions.

How the vulnerability chain worked

According to Cycode, the issue spans several vulnerability classifications: CWE-306 (missing authentication for critical functions), CWE-352 (cross-site request forgery, CSRF), and CWE-22 (path traversal). The AIT web server, by default, bound to the network interface 0.0.0.0 on port 8080, making it accessible from external networks without any restrictions. The endpoints for sending commands (/cmd), running scripts (/script/run), and sequences (/seq) did not require a password or login. Moreover, the system generated a session cookie sid for any anonymous request to the main page (GET /), which allowed bypassing the superficial session check and sending commands to the spacecraft. The /seq route additionally contained a Path Traversal vulnerability, opening the possibility of executing foreign files directly on the server.

Why the fix in version 2.5.2 is considered partial

The release of AIT-GUI version 2.5.2 limited the network listener to the local address localhost and added Origin/Referer header checks to protect against cross-site requests (CSRF). However, as Cycode researchers note, source code analysis showed that version 2.5.2 still did not add a basic authentication mechanism. This means that any request without credentials can still obtain a valid session cookie, and the web interface, designed to manage space missions, still does not ask for a password. The researchers classify the issue as only partially fixed, leaving an attack vector open for local or insider scenarios.

Contradictory data

Here arises a key discrepancy in assessments. On the one hand, the release of version 2.5.2 formally closes two of the three identified vectors — binding to 0.0.0.0 and CSRF attacks via a malicious web page in the operator's browser. On the other hand, the absence of authentication in the updated version means that the basic scenario — unauthorized sending of commands to a spacecraft via the local network or through a compromised host — remains active. Cycode explicitly states that the issue is "partially fixed," while the very fact of the patch release may create a false impression among administrators that the threat has been fully eliminated. Thus, the degree of closure of the vulnerability depends on whether it is viewed as a single chain or as a set of independent defects.

AI tools in vulnerability research

The research methodology deserves separate attention: in searching for the vulnerability, Cycode's cyber specialists used the Claude Opus 4.8 AI assistant from Anthropic, which helped analyze the AIT source code and write regression tests to verify the fixes. This case demonstrates the growing role of large language models in the security audit of critical infrastructure, including space systems.

Current status: no confirmed attacks recorded

As of the publication of the report, no cases of real-world exploitation of this vulnerability by attackers or compromise of active NASA space missions have been recorded. Nevertheless, given that AIT is open-source software and may be deployed not only in NASA's infrastructure but also in the systems of other space agencies and commercial operators, the risk of potential abuse remains relevant until a full authentication mechanism is implemented.