Deserialization of Untrusted Data Affecting nemo-toolkit package, versions [,2.6.2)

nemo-toolkit is a NeMo - a toolkit for Conversational AI

Affected versions of this package are vulnerable to Deserialization of Untrusted Data through the torch.load checkpoint and model import paths in the nemo collections and checkpoint utilities. An attacker can execute arbitrary code on a victim system by supplying a malicious PyTorch checkpoint or model file that the framework loads. The issue affects code paths that deserialize .ckpt, .nemo, and imported model weights during training, fine-tuning, checkpoint restoration, and model conversion.

Note: The vulnerable behavior is triggered when NeMo loads attacker-controlled serialized model artifacts from local paths or downloaded sources.

Serialization is a process of converting an object into a sequence of bytes which can be persisted to a disk or database or can be sent through streams. The reverse process of creating object from sequence of bytes is called deserialization. Serialization is commonly used for communication (sharing objects between multiple hosts) and persistence (store the object state in a file or a database). It is an integral part of popular protocols like Remote Method Invocation (RMI), Java Management Extension (JMX), Java Messaging System (JMS), Action Message Format (AMF), Java Server Faces (JSF) ViewState, etc.

Deserialization of untrusted data (CWE-502) is when the application deserializes untrusted data without sufficiently verifying that the resulting data will be valid, thus allowing the attacker to control the state or the flow of the execution.

• GitHub Commit
• NVIDIA Security Advisory