Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-85494 Overview
CVE-2026-85494 affects Apache Thrift across multiple language bindings including Python, Ruby, Erlang, Lua, Dart, JavaME, Perl, PHP, and D. The vulnerability combines several weaknesses: improper handling of length parameter inconsistency, uncaught exceptions, inefficient algorithmic complexity, memory allocation with excessive size values, and initialization of a resource with an insecure default. Attackers can trigger these conditions remotely over the network without authentication. All Apache Thrift releases before 0.25.0 are affected. The Apache Thrift project assigned this issue to CWE-130, covering improper handling of length parameter inconsistency.
Critical Impact
Unauthenticated remote attackers can disrupt availability of services built on Apache Thrift bindings by sending malformed messages that exhaust memory, trigger uncaught exceptions, or force expensive processing.
Affected Products
- Apache Thrift Python, Ruby, Erlang, Lua, Dart, JavaME, Perl, PHP, and D language bindings
- All versions prior to Apache Thrift 0.25.0
- Services and RPC endpoints built on the affected bindings
Discovery Timeline
- 2026-10-02 - CVE-2026-85494 published to NVD
- 2026-10-07 - Last updated in NVD database
Technical Details for CVE-2026-85494
Vulnerability Analysis
Apache Thrift is a cross-language framework for scalable services. The deserialization paths in several language bindings fail to validate length parameters embedded in incoming messages. An attacker can declare an oversized collection, string, or binary field, causing the receiver to allocate large buffers or iterate through fabricated lengths. The impact is availability loss: resource exhaustion, process termination from uncaught exceptions, or CPU saturation from algorithmic complexity. The flaw also includes an insecure default initialization in affected bindings, which lowers the barrier to triggering these conditions.
Root Cause
The root cause is tracked as CWE-130, improper handling of length parameter inconsistency. The affected bindings trust length fields in Thrift protocol messages without cross-checking them against the actual payload size or sane upper bounds. This single trust gap compounds into several observable failures: oversized memory allocation, uncaught exceptions when buffers cannot be filled, and quadratic processing paths when parsers attempt to reconcile malformed structures.
Attack Vector
The attack is network-reachable, requires no privileges, and no user interaction. An attacker sends a crafted Thrift message to any exposed service built on an affected binding. The malicious payload declares inflated lengths for list, map, set, string, or binary fields. On deserialization, the binding either allocates memory sized by the attacker-controlled length, raises an uncaught exception that terminates the handler, or enters an inefficient parsing path that consumes CPU. Repeated requests convert a single malformed message into a sustained denial-of-service condition.
No verified public exploit code was available at the time of writing. Review the Apache Discussion Thread and the Apache Mailing List Thread for protocol-level details.
Detection Methods for CVE-2026-85494
Indicators of Compromise
- Thrift service processes terminating with uncaught exceptions in Python, Ruby, Perl, PHP, Lua, Dart, Erlang, D, or JavaME runtimes
- Sudden memory allocation spikes correlated with inbound Thrift RPC traffic
- Elevated CPU utilization in deserialization code paths without corresponding business workload
- Repeated short-lived TCP sessions from a single source to Thrift listener ports
Detection Strategies
- Inspect Thrift protocol traffic for declared field lengths that exceed message size or configured service limits
- Monitor service restart loops and crash signatures referencing Thrift transport or protocol modules
- Correlate RPC error rates with memory and CPU metrics on hosts running Thrift servers
- Alert on deserialization exceptions originating from thrift libraries in application logs
Monitoring Recommendations
- Enable structured logging on Thrift servers to capture message size, declared field lengths, and parse failures
- Forward application and process-exit telemetry to a centralized data lake for correlation across hosts
- Track baseline request sizes per endpoint and alert on statistical outliers
How to Mitigate CVE-2026-85494
Immediate Actions Required
- Upgrade all Apache Thrift language bindings to version 0.25.0 or later
- Inventory services using Python, Ruby, Erlang, Lua, Dart, JavaME, Perl, PHP, and D Thrift bindings and prioritize internet-exposed endpoints
- Place rate limits and request-size caps in front of Thrift listeners until patches are deployed
- Restrict Thrift RPC endpoints to trusted network segments where feasible
Patch Information
The Apache Thrift project fixed this issue in version 0.25.0. Users of all affected language bindings should upgrade. Refer to the Apache Mailing List Thread for the official announcement and the Apache Discussion Thread for additional context.
Workarounds
- Configure maximum message size and container length limits on Thrift transports where the binding supports it
- Terminate client connections at a reverse proxy that enforces payload size ceilings before traffic reaches the Thrift server
- Isolate Thrift services behind authenticated gateways to reduce exposure to unauthenticated attackers
- Restart affected services under a supervisor that enforces memory ceilings to limit the blast radius of a single malicious request
# Example: cap request body size at an NGINX reverse proxy fronting a Thrift HTTP transport
http {
client_max_body_size 1m;
server {
listen 443 ssl;
location /thrift {
proxy_pass http://thrift_backend;
proxy_read_timeout 10s;
}
}
}
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

