Resilience of Pruned Neural Network Against Poisoning Attack

Zhao, Bingyin; Lao, Yingjie

doi:10.1109/malware.2018.8659362

Cited by 8 publications

(3 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There exists work [56,27] attempting to directly mitigate the backdoor attack without inspecting the model behavior. The key technique behind these methods is to compress the model (e.g., by model pruning or similar techniques) or fine-tune the model with benign inputs to alter the model behavior hoping that the backdoor behavior is eliminated.…”

Section: Pre-deployment Techniques No Inspectionsmentioning

confidence: 99%

“…The key technique behind these methods is to compress the model (e.g., by model pruning or similar techniques) or fine-tune the model with benign inputs to alter the model behavior hoping that the backdoor behavior is eliminated. Specifically, Zhao et al [56] found that model pruning can remove some behaviors of a trained model, and potentially it can remove the backdoor of the model if pruning is purely using benign data.…”

Section: Pre-deployment Techniques No Inspectionsmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning Backdoors

Li¹,

Ma²,

Xue³

et al. 2020

Preprint

View full text Add to dashboard Cite

Section: Pre-deployment Techniques No Inspectionsmentioning

confidence: 99%

Section: Pre-deployment Techniques No Inspectionsmentioning

confidence: 99%

Deep Learning Backdoors

Li¹,

Ma²,

Xue³

et al. 2020

Preprint

View full text Add to dashboard Cite

“…Finally, the authors devised a methodology for trojan detection by investigating the simulations of multiple types of embedded trojans. In comparison to previous work, the trojan detection approach is an extension of the observation in [ 12 ] about pruned NNs having a higher resilience against adding malicious triggers. Thus, two identical models, one with and one without embedded trojan, will have different inefficiency/utilization measured by the modified KL divergence.…”

Section: Introductionmentioning

confidence: 99%

Designing Trojan Detectors in Neural Networks Using Interactive Simulations

2021

View full text Add to dashboard Cite

This paper addresses the problem of designing trojan detectors in neural networks (NNs) using interactive simulations. Trojans in NNs are defined as triggers in inputs that cause misclassification of such inputs into a class (or classes) unintended by the design of a NN-based model. The goal of our work is to understand encodings of a variety of trojan types in fully connected layers of neural networks. Our approach is: (1) to simulate nine types of trojan embeddings into dot patterns; (2) to devise measurements of NN states; and (3) to design trojan detectors in NN-based classification models. The interactive simulations are built on top of TensorFlow Playground with in-memory storage of data and NN coefficients. The simulations provide analytical, visualization, and output operations performed on training datasets and NN architectures. The measurements of a NN include: (a) model inefficiency using modified Kullback–Liebler (KL) divergence from uniformly distributed states; and (b) model sensitivity to variables related to data and NNs. Using the KL divergence measurements at each NN layer and per each predicted class label, a trojan detector is devised to discriminate NN models with or without trojans. To document robustness of such a trojan detector with respect to NN architectures, dataset perturbations, and trojan types, several properties of the KL divergence measurement are presented.

show abstract

Deep Learning Backdoors

Ma²,

Xue

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Resilience of Pruned Neural Network Against Poisoning Attack

Cited by 8 publications

References 19 publications

Deep Learning Backdoors

Deep Learning Backdoors

Designing Trojan Detectors in Neural Networks Using Interactive Simulations

Deep Learning Backdoors

Contact Info

Product

Resources

About