The vulnerability arises from insufficient input sanitization in the Snowflake Python API (snowflake.core) in versions prior to 1.13.0. Two weaknesses can be exploited: a path traversal flaw where an unencoded '..' segment in an object name leads the API to issue REST calls against a parent resource or an HTTP parameter pollution flaw where unencoded '&', '#', or '=' characters in query string values override constraints on swap, clone, or rename operations. An attacker who can control an identifier or object‑name string in a downstream application built on snowflake.core can trigger these flaws, causing the API to execute operations under a higher‑privileged Snowflake session, thereby achieving privileged actions such as data manipulation or access to resources normally restricted to the session owner. The attack compromises both confidentiality and integrity by allowing unauthorized modification or exposure of data and effectiveness of the scope is limited to the privileges attached to the session in which the API call is made.

Snowflake Python APIs (snowflake.core) in all releases earlier than 1.13.0 are affected. Any application that utilizes snowflake.core, including Streamlit apps, Native Apps, or any stored procedures that execute under privileges such as EXECUTE AS OWNER, may be susceptible when untrusted input is passed to the API in an elevated session.

The CVSS base score of 8.1 indicates a high severity vulnerability, and while an EPSS score is not available, the lack of a KEV listing suggests no widely available commercial exploitation yet. Exploitation requires that an attacker can inject a malicious identifier or object name into a snowflake.core call that runs with elevated privileges. Because the flaw is in input handling, the likely attack vector is through an application that accepts user‑supplied names or identifiers and forwards them unchecked to snowflake.core, making this vulnerability a potential vector for confused‑deputy privilege escalation.