Backdoor attacks against learning systems

70 citations recorded by the source corpus.

How researchers see this paper

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

Model-Reuse Attacks on Deep Learning Systems (2018, background): “[30] consider a similar setting, but with the assumption that the adversary has access to the training data in both source and target domains.”
A New Backdoor Attack in CNNS by Training Set Corruption Without Label Poisoning (2019, background): “Most backdoor attacks assume that the model is fully or partially known to the attacker and under its control up to some extent [1], [2], [6], [7].”
ABS: Scanning Neural Networks for Back-doors by Artificial Brain Stimulation (2019, background): “There are also general poisoning attacks [28, 29, 51] where the attackermanipulates the training data tomake themodel ineffective.”
Backdoor Embedding in Convolutional Neural Network Models via Invisible Perturbation (2018, background): “bove 90% with an injection rate around 1%, and incur a loss in classification accuracy of less than 1%. Highlight Differences.The concept of a backdoor attack has been proposed by some recent studies [9, 17, 21, 33]. We draw a more detailed comparison with respect to three facets: assumptions on the”
Detecting AI Trojans Using Meta Neural Analysis (2019, background): “Several recent research [37], [22], [36], [14], [34], [26] has studied software Trojan attacks on neural networks.”