SiYuan CVE-2026-73048

The SiYuan instance must be network-accessible and must have at least one publicly published page that contains PDF annotations - annotation identifiers visible in those published pages serve as the required input to the vulnerable endpoint. The getRefIDsByFileAnnotationID endpoint requires no authentication, consistent with PR:N in the provided CVSS 4.0 vector, and works against default configurations when the public sharing feature is active. Exploitation produces meaningful intelligence only when the deployment operates mixed access tiers - some documents published and others restricted or password-protected - because the leak reveals citation relationships that cross those tier boundaries. SiYuan instances without public sharing enabled, or where all documents share a uniform access tier, present no exploitable attack surface from this specific vulnerability. What limits real-world impact: the disclosure is scoped to block identifiers and citation metadata, not document text content; and the vulnerability is only present in SiYuan installations prior to v3.7.4.

The CVSS 4.0 base score of 6.9, driven by a vector of AV:N/AC:L/AT:N/PR:N/UI:N, confirms remote exploitability with zero authentication and no complexity barrier. The impact metrics VC:L and SC:L reflect that only block identifiers and citation metadata are exposed, not full document content, keeping this in the medium-severity band rather than critical. No EPSS score was present in the provided intelligence, and this CVE does not appear in the CISA KEV catalog - exploitation probability must therefore be estimated from structural signals alone. The absence of a public POC keeps near-term automated exploitation risk low, but the trivially reachable endpoint means any attacker aware of the API can immediately begin extracting data with nothing more than an HTTP client. In practice, even a mid-sized municipal IT team in kalisz running SiYuan as a shared internal wiki and exposing it to the internet without perimeter controls would face real, concrete information-disclosure risk - citation relationships and knowledge-graph topology can themselves be sensitive, enabling targeted social engineering or revealing project associations. The primary limiting factor is SiYuan's niche deployment footprint and the requirement that the target instance simultaneously use public sharing and mixed access-tier documents.

An attacker visits a publicly shared SiYuan page and harvests PDF annotation identifiers that are embedded by design in the published content, then issues unauthenticated HTTP GET requests to the getRefIDsByFileAnnotationID endpoint, supplying those annotation IDs as parameters. The API responds with block identifiers drawn from restricted or password-protected documents, allowing the attacker to reconstruct citation relationships and confirm the existence and inter-referencing structure of documents they should never be able to access. No public exploit code has been identified at time of analysis, but the attack requires no specialized tooling beyond a standard HTTP client and knowledge of the API endpoint path.