Understanding the risk factors of learning in adversarial environments

By Blaine Nelson, Battista Biggio, Pavel Laskov. 2011

23 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Safety, Robustness & Interpretability

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

1. Battista Biggio
2. Fabio Roli
3. Giorgio Fumera
4. Nicolas Papernot
5. Claudia Eckert
6. Giorgio Giacinto
7. Huang Xiao
8. Blaine Nelson

In Computer Networks and Communications

1. Battista Biggio
2. Fabio Roli
3. Giorgio Fumera
4. Blaine Nelson
5. Claudia Eckert
6. Huang Xiao
7. Han Xiao
8. Brent Lagesse

In Safety, Robustness & Interpretability

1. Battista Biggio
2. Fabio Roli
3. Giorgio Fumera
4. Claudia Eckert
5. Huang Xiao
6. Nicolas Papernot
7. Giorgio Giacinto
8. Han Xiao

Citations in the corpus

• Wild Patterns: Ten Years After the Rise of Adversarial Machine Learning (2017, background): “While most work focused on countering evasion attacks at test time, some white-box defenses have also been proposed against poisoning attacks [22, 23, 94, 96–99].”
• Learning Discrete Distributions from Untrusted Batches (2017, background): “On the applied side, the interest in “ad-2 versarial” machine learning stems from the very real vulnerabilities of many deployed learning systems to “data poisoning” attacks in which an adversary selectively alters or plants a small amount of carefully chosen training data that significantly impacts ”
• Support vector machines under adversarial label contamination (2015, methodology): “However, instead of maximizing the test error directly, here we consider a surrogate measure, inspired by our work in [44].”
• Systematic Poisoning Attacks on and Defenses for Machine Learning in Healthcare (2015, background): “More recent efforts have considered the possibility that vulnerabilities in machine-learning algorithms may be exploited by attackers to influence the algorithm’s results [12]–[16].”
• Pattern Recognition Systems under Attack: Design Issues and Research Challenges (2014, background): “While most works focused on countering evasion attacks at test time (which do not involve manipulating the training data), some work dealt with the security of the training process in the presence of training data contamination, which may occur when the system is updated with data collected during o”
• Analysis of Causative Attacks against SVMs Learning from Data Streams (2017, background): “In [11], classifier robustness is calculated according to the impact that an adversary may have on a learning algorithm.”