Communications in computer and information science, Год журнала: 2024, Номер unknown, С. 263 - 278
Опубликована: Янв. 1, 2024
Язык: Английский
Communications in computer and information science, Год журнала: 2024, Номер unknown, С. 263 - 278
Опубликована: Янв. 1, 2024
Язык: Английский
Journal of Cybersecurity and Privacy, Год журнала: 2024, Номер 4(4), С. 972 - 992
Опубликована: Ноя. 17, 2024
In recent years, Machine Learning (ML) and Artificial Intelligence (AI) have been gaining ground in Cyber Security (CS) research an attempt to counter increasingly sophisticated attacks. However, this paper poses the question of qualitative quantitative data. This argues that scholarly domain is severely impacted by quality quantity available Datasets are disparate. There no uniformity (i) dataset features, (ii) methods collection, or (iii) preprocessing requirements enable good-quality analyzed data suitable for automated decision-making. review contributes existing literature providing a single summary wider field relation AI, evaluating most datasets, combining considerations ethical posing list open questions guide future endeavors. Thus, valuable insights cyber security field, fostering advancements application AI/ML.
Язык: Английский
Процитировано
2Mathematics, Год журнала: 2024, Номер 12(12), С. 1813 - 1813
Опубликована: Июнь 11, 2024
In online social networks, users can vote on different trust levels for each other to indicate how much they their friends. Researchers have improved ability predict relationships through a variety of methods, one which is the graph neural network (GNN) method, but also brought vulnerability GNN method into model. We propose data-poisoning attack GNN-based models based characteristics networks. used two-sample test power-law distributions discrete data avoid changes in dataset being detected and an enhanced surrogate model generate poisoned samples. further tested effectiveness our approach three real-world datasets compared it with two methods. The experimental results using show that effectively detection. metrics illustrate attack, stayed ahead methods all datasets. terms metrics, decreased accuracies attacked by 12.6%, 22.8%, 13.8%.
Язык: Английский
Процитировано
1Communications in computer and information science, Год журнала: 2024, Номер unknown, С. 263 - 278
Опубликована: Янв. 1, 2024
Язык: Английский
Процитировано
0