Feature Weighting for Improved Classifier Robustness

By A. Kolcz, Choon Hui Teo. 2009

141 citations recorded by the source corpus.

Subfields: Optimization & Training, Safety, Robustness & Interpretability

How researchers see this paper

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

In Optimization & Training

In Safety, Robustness & Interpretability

Citations in the corpus

Wild Patterns: Ten Years After the Rise of Adversarial Machine Learning (2017, background): “…(poisoning attacks) [22–31] and at test time (evasion attacks) [18– 20, 28, 32–37]; ( ii ) propose systematic methodologies for security evaluation of learning algorithms against such attacks [37– 41]; and ( iii ) design suitable defense mechanisms to mitigate these threats [11, 18, 33, 42–46].”
Evasion Attacks against Machine Learning at Test Time (2013, background): “, to avoid detection, spam emails are often modified by obfuscating common spam words or inserting words associated with legitimate emails [3, 9, 16, 19].”
Poisoning behavioral malware clustering (2014, background): “Such a constraint can be often encoded by a suitable distance measure between the original, non-manipulated attack samples and the manipulated ones, as in [2, 7, 17, 23].”
Security Evaluation of Pattern Classifiers under Attack (2014, neutral): “F”
Yes, Machine Learning Can Be More Secure! A Case Study on Android Malware Detection (2017, methodology): “tic approaches in [8], [26], but it is also able to outperform them in terms of security.”
On Security and Sparsity of Linear Classifiers for Adversarial Settings (2016, background): “To evaluate security of linear classifiers, we define a measure E of weight evenness, similarly to [4,7], based on the ratio of the `1 and `∞ norm:”
Adversarial Feature Selection Against Evasion Attacks (2016, methodology): “It has been, therefore, widely adopted in previous work [10], [11], [15], [16], [23], [26], and in several real anti-spam filters, like SpamAssassin and SpamBayes.”
Security Evaluation of Support Vector Machines in Adversarial Environments (2014, background): “Although potentially too pessimistic, this worst-case setting allows one to compute a lower bound on the classifier performance when it is under attack [26, 41].”
Support vector machines under adversarial label contamination (2015, background): “Keywords: Support Vector Machines, Adversarial Learning, Label Noise, Label Flip Attacks”
On Evaluating Adversarial Robustness (2019, background): “This long line of research (Dalvi et al., 2004; Lowd & Meek, 2005; Barreno et al., 2006; 2010; Globerson & Roweis, 2006; Kołcz & Teo, 2009; Barreno et al., 2010; Biggio et al., 2010; Šrndic & Laskov, 2013) has recently begun seeing significant study as machine learning becomes more widely used.”
Secure Kernel Machines against Evasion Attacks (2016, background): “Adding or removing a term amounts to switching the value of the corresponding Boolean feature [4, 7, 18, 22, 36].”
Poisoning Complete-Linkage Hierarchical Clustering (2014, background): “Such a constraint can be expressed in terms of a suitable distance measure between the original, non-manipulated attack samples and the manipulated ones, as in [8, 17, 10, 11].”
Towards Deep Learning Models Resistant to Large Perturbations (2020, background): “Adversarial machine learning has been studied since two decades ago (Dalvi et al., 2004; Globerson & Roweis, 2006; Kolcz & Teo, 2009).”
Multiple Classifier Systems under Attack (2010, methodology): “Interestingly, it turns out that the effect of the proposed algorithm is similar to the strategy of [7], namely it results in producing more uniform weight values.”
Sparse Feature Attacks in Adversarial Learning (2014, methodology): “Similarly, another work uses feature reweighting to overcome the discrepancies between training and test data [5], where each feature vector is re-scaled by its length (i.”
Towards Adversarial Malware Detection (2018, background): “, in spam filtering [11, 50] and Android malware detection [28]) that learning algorithms may tend to overemphasize some features, and that this facilitates evasion by a skilled attacker who carefully manipulates only those feature values.”
Support Vector Machines Under Adversarial Label Noise (2011, background): “Due to their ability to generalize, machine learning algorithms have been widely adopted in security applications as well; e.g., in spam filtering and intrusion detection.”