The vulnerability exists in FedML’s gRPC server component, specifically in the sendMessage function of grpc_server.py. An attacker who can send a crafted message to the server can trigger unsafe deserialization of untrusted data. The flaw stems from improper input validation and the deserialization of potentially malicious payloads, which may allow the attacker to execute arbitrary code, compromising confidentiality, integrity or availability of the affected system.

FedML-AI’s FedML product, versions up to 0.8.9, includes the vulnerable gRPC server. Any installation that uses the sendMessage function in grpc_server.py before or at this version is at risk. The product is commonly used in distributed machine learning environments where remote connections to the gRPC service are required.

The CVSS base score is 6.9, indicating medium severity. No EPSS data is available and the vulnerability is not listed in CISA’s KEV catalog. The injection can be performed remotely, relying on network traffic to the gRPC port. Because the flaw involves deserialization, without additional defenses an attacker with network access could possibly chain the vulnerability into remote code execution, making the risk significant for exposed deployments.