Double Free

tensorflow is vulnerable to a Double Free attack. The vulnerability occurs when when the first and the fourth elements of the pooling_ratio function has parameters not equal to 1.0 in nn_ops.fractional_avg_pool_v2 and nn_ops.fractional_max_pool_v2 because pooling on batch and channel dimensions is not supported. An attacker can cause a Denial of Service or possibly alter the execution flow.