Vulnerabilities OFFIS DCMTK Toolkit Can Result in Data Deletion
Two more vulnerabilities have been identified in the OFFIS DCMTK toolkit that could result in the deletion of medical imaging data. The most serious vulnerability is tracked as CVE-2026-125130 and has a CVSS v3.1 severity score of 9.1. The vulnerability is due to a read-only archive area accepting network stores. Each unauthorized store can trigger the archive’s own quota cleanup, which would delete existing studies, potentially resulting in total data deletion from the archive. The vulnerability can be exploited by a network peer without credentials.
The vulnerabilities were identified by security researcher Abhinav Agarwal, who demonstrated in tests on DCMTK 3.7.0 that the server accepted 200 unauthorized stores, triggering the archive cleanup resulting in the deletion of 200 original synthetic image files. The final query confirmed that all 200 original image files had been deleted. Agarwal determined that while DCMTK enforced read-only access, that only occurred when a connection was set up. “If a peer asked to use the storage service on a read-only area, the server refused. It still accepted the Verification service (DICOM “echo”), which is harmless by itself,” explained Agarwal. “Once connected, however, the server executed whatever command arrived without checking that it matched the service negotiated for that connection. A peer could negotiate Verification only and then send storage commands over it.” Data deletion occurs when the read-only area is full, resulting in the deletion of the least recent imaging data to provide space for the most recent store.
The second vulnerability is tracked as CVE-2026-126500 and has a CVSS v3.1 score of 5.9 (medium severity). The vulnerability is due to the cleanup process deleting studies by prefix, rather than using exact matching. “To evict a study, cleanup picks the oldest one, then deletes every index record whose Study Instance UID starts with the selected UID,” said Agarwal. “Study selection elsewhere in the same backend uses exact matching. Cleanup did not.”
Agarwal demonstrated in tests that evicting the study ending .1 also deleted the studies ending .10–.19 and .100–.199 – resulting in 111 studies deleted in a single eviction. Data deletion could occur without any attacker involvement and could result in data loss from unrelated studies. Agarwal also uncovered five unrelated vulnerabilities in the same toolkit earlier this year, which have since been fixed. The latest two vulnerabilities were reported to the developer and CISA and affect versions 3.5.4 to 3.7.0. The developer fixed both vulnerabilities in DCMTK version 3.7.1, which was released on October 15, 2026.
While the new version fixes the vulnerabilities, they may still be present in products that incorporate the DCMTK archive server, including certain Debian, Ubuntu, and Fedora package sources. “An SBOM line that says ‘DCMTK’ doesn’t answer the question. Hospitals and suppliers need to know whether the archive server or its backend is in the build, which revision it came from, and whether both fixes made it downstream”, said Agarwal.”
