FADER: Fast Adversarial Example Rejection

no code implementations • 18 Oct 2020 • Francesco Crecchi, Marco Melis, Angelo Sotgiu, Davide Bacciu, Battista Biggio

As a second main contribution of this work, we introduce FADER, a novel technique for speeding up detection-based methods.

Adversarial Robustness

Perplexity-free Parametric t-SNE

1 code implementation • 3 Oct 2020 • Francesco Crecchi, Cyril de Bodt, Michel Verleysen, John A. Lee, Davide Bacciu

The t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm is a ubiquitously employed dimensionality reduction (DR) method.

Dimensionality Reduction

Detecting Adversarial Examples through Nonlinear Dimensionality Reduction

1 code implementation • 30 Apr 2019 • Francesco Crecchi, Davide Bacciu, Battista Biggio

Deep neural networks are vulnerable to adversarial examples, i. e., carefully-perturbed inputs aimed to mislead classification.

Density Estimation Dimensionality Reduction +1

DropIn: Making Reservoir Computing Neural Networks Robust to Missing Inputs by Dropout

1 code implementation • 7 May 2017 • Davide Bacciu, Francesco Crecchi, Davide Morelli

The paper presents a novel, principled approach to train recurrent neural networks from the Reservoir Computing family that are robust to missing part of the input features at prediction time.