vllm inherits pytorch s torch.distributed methodologies that support the distributed training over different cpu/gpu resources over networks (i.e. by distributed_init_method= tcp://ip:port ). Among the many functionalities in the distributed training, I found the function vllm.distributed.GroupCoordinator.recv_object() applies a pickle.loads() to deserialize the received object s bytes without any sanitization, hence resulting in a remote code execution vulnerability by this function.
vllm inherits pytorch s torch.distributed methodologies that support the distributed training over different cpu/gpu resources over networks (i.e. by distributed_init_method= tcp://ip:port ). Among the many functionalities in the distributed training, I found the function vllm.distributed.GroupCoordinator.recv_object() applies a pickle.loads() to deserialize the received object s bytes without any sanitization, hence resulting in a remote code execution vulnerability by this function.
vllm inherits pytorch s torch.distributed methodologies that support the distributed training over different cpu/gpu resources over networks (i.e. by distributed_init_method= tcp://ip:port ). Among the many functionalities in the distributed training, I found the function vllm.distributed.GroupCoordinator.recv_object() applies a pickle.loads() to deserialize the received object s bytes without any sanitization, hence resulting in a remote code execution vulnerability by this function.
vllm inherits pytorch s torch.distributed methodologies that support the distributed training over different cpu/gpu resources over networks (i.e. by distributed_init_method= tcp://ip:port ). Among the many functionalities in the distributed training, I found the function vllm.distributed.GroupCoordinator.recv_object() applies a pickle.loads() to deserialize the received object s bytes without any sanitization, hence resulting in a remote code execution vulnerability by this function..
vllm inherits pytorch s torch.distributed methodologies that support the distributedtrainingover different cpu/gpu resources over networks (i.e. by distributed_init_method= tcp://ip:port ).Among the manyfunctionalities in the distributedtraining, Ifound thefunction vllm.distributed.GroupCoordinator.recv_object() applies a pickle.loads() to deserialize the received object s bytes without any sanitization, hence resulting in a remote code execution vulnerability by this function..
vllm-project vllm version 0.6.0 contains a vulnerability in the distributed trainingAPI. Thefunction vllm.distributed.GroupCoordinator.recv_object() deserializes received object bytes usingpickle.loads()without sanitization, leading to aremote codeexecutionvulnerability.
vllm-project vllm version0.6.0 contains a vulnerability in the distributed training API. The function vllm.distributed.GroupCoordinator.recv_object() deserializes received object bytes using pickle.loads() without sanitization, leading to a remote code execution vulnerability.
