CVE-2025-66516

Apache Tika is vulnerable to XML External Entity (XXE) injection (CWE-611), rated critical at CVSS 9.8 (network-exploitable, no privileges or user interaction required), with the vulnerable code residing in tika-core but reachable through the PDF parser when it processes the XFA form embedded in a PDF. An attacker who can get Tika to parse a crafted PDF, typically by submitting a document to a Tika-based service or tika-server endpoint, can have the XML parser resolve external entities, gaining arbitrary local file read and SSRF from the Tika process, with CVSS impact rated high for confidentiality, integrity and availability. All platforms are affected across tika-core 1.13–3.2.1, tika-parser-pdf-module 2.0.0–3.2.1, and the 1.x tika-parsers module 1.13–1.28.5; notably the fix is in tika-core (3.2.2+), so upgrading only the PDF module leaves deployments vulnerable, and 1.x users need fixes in the tika-parsers module. Anyone running affected Tika versions to process untrusted PDFs is exposed, including standalone tika-server instances and downstream applications that embed Tika (vendors such as Atlassian have shipped fixes for embedded Tika). There is no CISA KEV listing and no public PoC or confirmed in-the-wild exploitation known, but EPSS puts it in the 100th percentile with an 80.3% probability of exploitation within 30 days.

What to do: Upgrade tika-core to 3.2.2 or later (upgrading only tika-parser-pdf-module is insufficient), and keep tika-parser-pdf-module aligned on the fixed 3.x line; users on the 1.x line should upgrade both tika-parsers and tika-core to the patched 1.x releases. If you consume Tika through another product (e.g., Solr-based stacks or Atlassian products), apply that vendor's patch. Until patched, avoid parsing untrusted PDFs with affected Tika versions, and restrict the Tika process's file-system and network access to limit file-read and SSRF impact.