How Vulnerable Are Automatic Fake News Detection Methods to Adversarial Attacks?