Summary of Outlier-oriented Poisoning Attack: a Grey-box Approach to Disturb Decision Boundaries by Perturbing Outliers in Multiclass Learning, By Anum Paracha and Junaid Arshad and Mohamed Ben Farah and Khalid Ismail
Outlier-Oriented Poisoning Attack: A Grey-box Approach to Disturb Decision Boundaries by Perturbing Outliers in Multiclass Learning
by Anum Paracha, Junaid Arshad, Mohamed Ben Farah, Khalid Ismail
First submitted to arxiv on: 1 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel Outlier-Oriented Poisoning (OOP) attack manipulates labels of most distanced samples from decision boundaries in machine learning models, compromising performance and reliability. This paper investigates the OOP attack’s impact on multiclass classification algorithms, analyzing variance, accuracy, precision, recall, F1-score, and false positive rate across different poisoning levels (5%-25%). The study used three publicly available datasets: IRIS, MNIST, and ISIC, with KNN and GNB being the most affected algorithms. Decision Trees and Random Forest were found to be more resilient, while the number of dataset classes is inversely proportional to performance degradation. Imbalanced dataset distribution can exacerbate the attack’s impact. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning models are vulnerable to attacks that manipulate training datasets, compromising their performance and reliability. A new type of attack, called Outlier-Oriented Poisoning (OOP), targets the most distant samples from decision boundaries, making it harder for models to learn. This paper looks at how OOP affects different machine learning algorithms and what happens when you mix in some imbalanced data. The results show that some algorithms are more resistant than others and that having too many classes can help mitigate the attack’s impact. |
Keywords
* Artificial intelligence * Classification * F1 score * Machine learning * Precision * Random forest * Recall
