Regarding duplicated queries in the Square Attack

[Lodour]

Is your feature request related to a problem? Please describe. I was benchmarking the performance of some black-box attacks and noticed that the Square Attack issued about 50% duplicated queries. This occurs because most queries are sent twice in a subtle way; see the code below.

Describe the solution you'd like Update the implementation to reuse the last prediction outputs.

I am happy to send a PR if you find this performance improvement useful.

Describe alternatives you've considered N/A

Additional context For example, x_adv is passed to self.estimator.predict at L347, and its subset x_robust = x_adv[sample_is_robust] is immediately passed to self.loss, which boils down to a duplicated call of self.estimator.predict on the same inputs. https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/337a15f9db8e447426aa9cbe1f090d14e3f1d3a4/art/attacks/evasion/square_attack.py#L346-L357

[beat-buesser]

Hi @Lodour Thank you for using ART! How does you proposed solution with the same functionality look like?

[Lodour]

My workaround is passing y_pred to self.loss, but this only handles the default loss function self._get_logits_diff

https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/337a15f9db8e447426aa9cbe1f090d14e3f1d3a4/art/attacks/evasion/square_attack.py#L118-L126

The method would become something like

def _get_logits_diff(self, x: np.ndarray, y: np.ndarray, y_pred: Optional[np.ndarray] = None):
    if y_pred is None:
        y_pred = self.estimator.predict(x, batch_size=self.batch_size)
    ...

I haven't come up with a general solution for custom loss functions.