Apache Thrift EUVD-2026-91560

Exploitation requires a service that uses the Apache Thrift Python bindings (version before 0.25.0) and accepts externally supplied Thrift data that is decompressed or size-expanded during parsing; the AT:P metric reflects that the attacker needs the target to process such compressed/amplified input. No authentication is required (PR:N) and no user interaction (UI:N). Limiting factors: only services actually exposing a Thrift Python endpoint to untrusted input are affected, impact is confined to availability (no data disclosure or code execution), and upgrading to 0.25.0 removes the condition entirely.

The CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:N/UI:N) with score 8.2 describes a network-reachable, unauthenticated, low-complexity flaw whose only impact is availability (VA:H; VC:N/VI:N/VA:H), consistent with CWE-409 data-amplification (decompression/entity-expansion bomb) in the Apache Thrift Python bindings. The AT:P (attack requirements present) metric indicates exploitation is not fully unconditional - the target endpoint must accept and decompress attacker-supplied data - which tempers the otherwise high exposure. No EPSS score, CISA KEV listing, or public POC was provided in the input, so there is no evidence of active exploitation or weaponization: this reads as a genuine but availability-only DoS priority rather than a confirmed in-the-wild threat. One tagging inconsistency should be noted explicitly: the input carries an 'Information Disclosure' tag, yet the CVSS vector sets VC:N/VI:N and only VA:H, meaning this is a denial-of-service/availability issue with no confidentiality impact - the tag appears mislabeled and should not drive triage toward data-exposure concerns.

An attacker sends a small but highly compressed or deeply nested payload (a 'bomb') to a network-exposed service built on the Apache Thrift Python bindings; when the library decompresses or expands the input without validating the declared size/quantity, memory and CPU are exhausted and the service becomes unavailable. Because the endpoint is reachable over the network without authentication, a single malicious request can degrade or crash the service. No public exploit code was identified at the time of analysis.