Making machine learning robust against adversarial inputs

By Ian J. Goodfellow, Patrick Drew McDaniel, Nicolas Papernot. Communications of the ACM, 2018

374 citations recorded by the source corpus.

Such inputs distort how machine-learning-based systems are able to function in the world as it is.

Subfields: Artificial Intelligence, Signal Processing

View Paper: https://www.semanticscholar.org/paper/e96fe4e820c2ca9002ad56c6a58c1e3840130fa7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Signal Processing

Citations in the corpus

Stakeholders in Explainable AI (2018, background): “…to model inter-pretability (Leino et al. 2018; Sundararajan, Taly, and Yan 2017), exploring more sophisticated methods for revealing the inner workings of deep networks (Olah et al. 2018), and arguing for increased use of theoretical veriﬁcation techniques (Goodfellow, McDaniel, and Papernot 2018).”
Make Up Your Mind! Adversarial Generation of Inconsistent Natural Language Explanations (2019, background): “To this date, most of the research on adversarial examples in computer vision focuses on generating adversarial perturbations that are imperceptible to humans, but make the machine learning model to produce a different prediction [12].”
Procedural Noise Adversarial Examples for Black-Box Attacks on Deep Convolutional Networks (2018, background): “oss function using gradient descent. Theoretically, gradient-based algorithms are more difﬁcult for neural networks on large-image datasets because these models use nonlinear and non-convex functions [43]. The gradients can get drawn to local optima as the search space and number of dimensions becom”
The Curse of Concentration in Robust Learning: Evasion and Poisoning Attacks from Concentration of Measure (2018, background): “uch as image classiﬁcation, in which the adversarially perturbed instance x′would still “look similar” to the original x, at least in humans’ eyes, even though the classiﬁer h might now misclassify x′[GMP18]. In fact, starting with the work of Szegedy et al. [SZS+14] an active line of research (e.g.”
A Survey on Machine Learning Techniques for Cyber Security in the Last Decade (2020, background): “A model Goodfellow [658] was claimed to be robust against adversarial inputs.”
Lightlike Neuromanifolds, Occam's Razor and Deep Learning (2019, background): “This robustness term is closely related to phenomena in deep learning, such as the ﬂatness of local optima [9] and adversary learning [14], and inspires us to apply eq.”
A decade of social bot detection (2020, background): “Generative adversarial networks (GANs) are a powerful machine learning framework where two competing deep learning networks are jointly trained in a game-theoretic setting [15].”