Denial Of Service (DoS)

tensorflow is vulnerable to denial of service. A heap underflow occurs because invalid values for axis argument are accepted in tf.raw_ops.QuantizeAndDequantizeV2, allowing an attacker to read/write to other data on the heap.