Taxonomy of Machine Learning Safety: A Survey and Primer

Mohseni, Sina; Wang, Haotao; Yu, Zhiding; Xiao, Chaowei; Wang, Zhangyang; Yadawa, Jay

doi:10.48550/arxiv.2106.04823

Cited by 5 publications

(4 citation statements)

References 183 publications

(257 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Safe navigation is also a priority for applications. Navigation should avoid certain states if they are unethical to create, or if their creation would have negative ecological consequences (Aswani et al, 2013;Mohseni et al, 2021). Notably, many paths discovered by our approach transiently put the community into higher richness states that include novel species (e.g., orangecoloured states in Figure 2).…”

Section: Extensions To the Navigation Approachmentioning

confidence: 96%

Navigation between initial and desired community states using shortcuts

et al. 2023

View full text Add to dashboard Cite

Ecological management problems often involve navigating from an initial to a desired community state. We ask whether navigation without brute-force additions and deletions of species is possible via: adding/deleting a small number of individuals of a species, changing the environment, and waiting. Navigation can yield direct paths (single sequence of actions) or shortcut paths (multiple sequences of actions with lower cost than a direct path). We ask (1) when is non-brute-force navigation possible?; (2) do shortcuts exist and what are their properties?; and(3) what heuristics predict shortcut existence? Using a state diagram framework applied to several empirical datasets, we show that (1) non-brute-force navigation is only possible between some state pairs, (2) shortcuts exist between many state pairs; and (3) changes in abundance and richness are the strongest predictors of shortcut existence, independent of dataset and algorithm choices. State diagrams thus unveil hidden strategies for manipulating species coexistence and efficiently navigating between states.

show abstract

Section: Extensions To the Navigation Approachmentioning

confidence: 96%

Navigation between initial and desired community states using shortcuts

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Other techniques mainly include clustering, dataset analysis techniques, prediction techniques, estimation techniques (3%), association techniques, and statistical analysis. The study showed that public datasets accounted for 79% of experiments and experiments on private datasets were 21%, which were compared in the research studies [25]. A comprehensive study of 18 techniques for intrusion detection was conducted: six of these methods were found to be most prevalent: support vector machine, deep neural network, random forest, naive Bayes, decision tree, and K-nearest neighbor.…”

Section: Literature Reviewmentioning

confidence: 99%

Explainable Artificial Intelligence for Intrusion Detection System

et al. 2022

View full text Add to dashboard Cite

Intrusion detection systems are widely utilized in the cyber security field, to prevent and mitigate threats. Intrusion detection systems (IDS) help to keep threats and vulnerabilities out of computer networks. To develop effective intrusion detection systems, a range of machine learning methods are available. Machine learning ensemble methods have a well-proven track record when it comes to learning. Using ensemble methods of machine learning, this paper proposes an innovative intrusion detection system. To improve classification accuracy and eliminate false positives, features from the CICIDS-2017 dataset were chosen. This paper proposes an intrusion detection system using machine learning algorithms such as decision trees, random forests, and SVM (IDS). After training these models, an ensemble technique voting classifier was added and achieved an accuracy of 96.25%. Furthermore, the proposed model also incorporates the XAI algorithm LIME for better explainability and understanding of the black-box approach to reliable intrusion detection. Our experimental results confirmed that XAI LIME is more explanation-friendly and more responsive.

show abstract

“…Other representative benchmark efforts include [39][40][41][42]3], all demonstrating the brittleness of deep models under various distribution shifts. Readers of interest are referred to a recent survey [43].…”

Section: Robustness To Distributional Shiftsmentioning

confidence: 99%

AugMax: Adversarial Composition of Random Augmentations for Robust Training

Wang

Xiao

Kossaifi

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Data augmentation is a simple yet effective way to improve the robustness of deep neural networks (DNNs). Diversity and hardness are two complementary dimensions of data augmentation to achieve robustness. For example, AugMix explores random compositions of a diverse set of augmentations to enhance broader coverage, while adversarial training generates adversarially hard samples to spot the weakness. Motivated by this, we propose a data augmentation framework, termed AugMax, to unify the two aspects of diversity and hardness. AugMax first randomly samples multiple augmentation operators and then learns an adversarial mixture of the selected operators. Being a stronger form of data augmentation, AugMax leads to a significantly augmented input distribution which makes model training more challenging. To solve this problem, we further design a disentangled normalization module, termed DuBIN (Dual-Batch-and-Instance Normalization), that disentangles the instance-wise feature heterogeneity arising from AugMax. Experiments show that AugMax-DuBIN leads to significantly improved out-of-distribution robustness, outperforming prior arts by 3.03%, 3.49%, 1.82% and 0.71% on CIFAR10

show abstract

Taxonomy of Machine Learning Safety: A Survey and Primer

Cited by 5 publications

References 183 publications

Navigation between initial and desired community states using shortcuts

Navigation between initial and desired community states using shortcuts

Explainable Artificial Intelligence for Intrusion Detection System

AugMax: Adversarial Composition of Random Augmentations for Robust Training

Contact Info

Product

Resources

About