OneFlow Denial-of-Service Vulnerability in Tensor Input Validation

A denial-of-service vulnerability has been identified in OneFlow version 0.9.0. The issue arises in the input validation process of the flow.Tensor.new_empty, flow.Tensor.new_ones, and flow.Tensor.new_zeros methods. Attackers can exploit this vulnerability by providing crafted inputs that violate the methods' requirements, leading to application crashes.

Impact

Exploitation of this vulnerability causes the OneFlow application to crash, terminating the process and potentially disrupting any ongoing computations or services.

Reproduction

The vulnerability can be reproduced by calling the flow.Tensor.new_empty, flow.Tensor.new_ones, or flow.Tensor.new_zeros methods with negative dimensions. This input violates the non-negative requirement and triggers internal checks that, when failed, cause the application to crash. The issue can be observed by importing OneFlow, creating a tensor, and then using these methods with invalid input, which results in a core dump error.