SiYuan CVE-2026-72803

The vulnerable endpoints are part of SiYuan's HTTP API, which is network-accessible without authentication per the CVSS 4.0 PR:N rating - meaning no session, token, or user account is required to issue the POST requests. The primary limiting factor is block ID knowledge: SiYuan block IDs are application-internal identifiers that are not trivially guessable and must be obtained through prior exposure (e.g., shared links, exported content, or same-network sniffing). Exploitation is further bounded to attribute metadata (names, aliases, memos, custom fields) and does not yield document body text or attachments. Instances that are not internet-facing and are protected by network-layer controls are not practically exploitable without a foothold on the local network. The vulnerability is not conditional on any non-default configuration - any SiYuan instance running a version before 3.7.4 with network reachability is affected.

The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N) places this vulnerability at high network reach with minimal attack prerequisites, which warrants attention. However, the confidentiality impact is capped at Low (VC:L) because only block attributes - not document body content - are exposed; this substantially limits data sensitivity unless custom fields contain secrets such as API keys or credentials. No EPSS data was provided in the intelligence feed, no CISA KEV listing exists, and no public proof-of-concept has been identified, all pointing to low current exploitation pressure. A significant practical constraint is that an attacker must supply valid block IDs, which are application-internal identifiers not easily enumerated without prior access or information leakage - this raises effective attack complexity beyond what the CVSS Low/None metrics formally capture. Deployments fully isolated behind a VPN or firewall with no internet exposure carry materially lower risk; internet-exposed SiYuan instances should treat this as higher priority.

An attacker with network access to an exposed SiYuan instance crafts POST requests directly to /api/attr/getBlockAttrs or /api/attr/batchGetBlockAttrs, supplying block IDs obtained through prior reconnaissance - such as shared document links, browser history, or API enumeration. Without any session token or credential, the server returns attribute metadata for blocks inside documents the owner designated as protected, potentially including aliases, memos, and custom key-value fields that the owner assumed were private. No public proof-of-concept has been identified at time of analysis.