EUVD-2026-53000
GHSA-v276-wxc4-7v7p
Severity by source
CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:P/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
Local attack vector and low privileges required to submit filter queries; only availability impacted via parser crash, no confidentiality or integrity effect confirmed.
Primary rating from Vendor (VulDB).
CVSS VectorVendor: VulDB
Lifecycle Timeline
4DescriptionCVE.org
A vulnerability has been found in epsilla-cloud vectordb up to 0.3.18/df5a5f5afb85a2376a0f2f316c79dea9b2c6ac7a. This impacts the function SplitTokens/ShuntingYard of the file engine/query/expr/expr.cpp of the component Filter Parser. Such manipulation leads to improper check for unusual conditions. The attack needs to be performed locally. The exploit has been disclosed to the public and may be used. The vendor was contacted early about this disclosure but did not respond in any way.
AnalysisAI
Epsilla-cloud VectorDB's filter query parser crashes under malformed input due to an improper exception/condition check (CWE-754) in the SplitTokens and ShuntingYard functions of engine/query/expr/expr.cpp. All releases from 0.3.0 through 0.3.18 (including commit df5a5f5afb85a2376a0f2f316c79dea9b2c6ac7a) are affected. A local, low-privileged user who can submit filter expressions to the database engine can trigger an abnormal termination of the parser, causing an availability impact. A public proof-of-concept exploit exists; the vendor did not respond to responsible disclosure.
Technical ContextAI
VectorDB is a vector similarity search database engine by epsilla-cloud (CPE: cpe:2.3:a:epsilla-cloud:vectordb:*:*:*:*:*:*:*:*). The affected code path implements filter expression parsing using the Shunting-Yard algorithm - a well-known operator-precedence parsing technique - paired with a SplitTokens tokeniser, both residing in engine/query/expr/expr.cpp. CWE-754 (Improper Check for Unusual or Exceptional Conditions) indicates the parser does not correctly handle edge-case or malformed input tokens: rather than returning a controlled error, it proceeds into undefined or crash-inducing state. This class of bug commonly arises when expression parsers assume well-formed input and omit bounds or state validation on unexpected token sequences.
RemediationAI
No vendor-released patch has been identified at the time of analysis - the vendor did not respond to disclosure, and no fixed release is documented in the available references. As a compensating control, restrict local shell or API access to the VectorDB instance to trusted, explicitly authorized accounts only; reducing the pool of users who can submit arbitrary filter expressions directly limits exposure. Input validation or a query sanitisation wrapper applied at the application layer before expressions reach the filter parser can prevent malformed tokens from reaching the vulnerable code path. Operators should monitor the VulDB advisory at https://vuldb.com/vuln/385857 and the researcher's GitHub advisory at https://github.com/fa1c4/security-advisories/tree/main/VectorDB for any upstream patch. If the instance is exposed to untrusted local users, consider upgrading to an alternative vector database until a patched release is available.
Same technique Information Disclosure
View allShare
External POC / Exploit Code
Leaving vuln.today