A framework for quantitative security analysis of machine learning

By Pavel Laskov, Marius Kloft. 2009

76 citations recorded by the source corpus.

Subfields: Safety, Robustness & Interpretability, Privacy & Security

How researchers see this paper

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

In Safety, Robustness & Interpretability

In Privacy & Security

Citations in the corpus

Using Machine Teaching to Identify Optimal Training-Set Attacks on Machine Learners (2015, background): “Introduction The study on security threats to intelligent agents has a long history (Nelson et al. 2009; Tan, Killourhy, and Maxion 2002; Rubinstein et al. 2008; Barreno et al. 2010; Laskov and Lippmann 2010; Barreno et al. 2006; Dalvi et al. 2004; Liu and Chawla 2009; Biggio, Fumera, and Roli 2013;”
Evasion Attacks against Machine Learning at Test Time (2013, background): “As suggested by Laskov and Kloft [17], the adversary’s goal should be deﬁned in terms of a utility (loss) function that the adversary seeks to maximize (minimize).”
Security Evaluation of Pattern Classifiers under Attack (2014, neutral): “F”
Security Evaluation of Support Vector Machines in Adversarial Environments (2014, methodology): “Further, as suggested by Laskov and Kloft [42] and Kloft and Laskov [40], the adversary’s goal should be defined in terms of an objective function that the adversary is willing to maximize.”
Security analysis of online centroid anomaly detection (2010, methodology): “The methodology of our analysis follows the following framework, which we believe can be used for a quantitative security analysis of learning algorithms (Laskov and Kloft, 2009): 1.”
Data Poisoning Attacks against Autoregressive Models (2016, methodology): “The security ramifications of using machine learning models have long been studied (Barreno et al. 2006; 2010; Dalvi et al. 2004; Laskov and Kloft 2009; Tan, Killourhy, and Maxion 2002).”
Attacks on Machine Learning: Lurking Danger for Accountability (2019, unknown): “Laskov and Kloft [38] propose a “framework for quantitative security analysis of ML models”.”
Some Submodular Data-Poisoning Attacks on Machine Learners (2015, background): “The security community has long recognized the threats of data-poisoning attacks (a.k.a. causative attacks) on machine learning systems [1–6,9,10,12,16], where an attacker modiﬁes the training data, so that the learning algorithm arrives at a “wrong” model that is useful to the attacker.”
Behavior of Machine Learning Algorithms in Adversarial Environments (2010, background): “Laskov and Kloft [2009] have since extended these basic principles to propose a framework for quantitatively evaluating security threats.”
Adversarial machine learning (2019, methodology): “In prior work [2], we introduced a qualitative taxonomy of attacks against a machine learning system, which has since been extended to propose a framework for quantitatively evaluating security threats [37].”
The relationship between trust in AI and trustworthy machine learning technologies (2019, background): “Integrity can be enhanced by either preventing tampering, such as in pre-processing [73], or by discovering and possibly repairing tampered data [12, 14, 58].”
Multiple classifier systems for robust classifier design in adversarial environments (2010, background): “Evaluating classifier robustness in adversarial classification tasks is a complex issue, which is outside the scope of this paper, and is the subject of an ongoing work [13,37], also by the authors.”
Identifying Security Evaluation of PatternClassifiers Under attack (2015, background): “An intrinsic limitation of our work is that security evaluation is carried out empirically, and it is thus data dependent; on the other hand, model-driven analyses [12], [17], [38] require a full analytical model of the problem and of the adversary’s behavior, that may be very difficult to develop f”
Detection of Adversarial Training Examples in Poisoning Attacks through Anomaly Detection (2018, background): “However, machine learning systems are vulnerable themselves [4]–[10], and can be targeted by attackers.”
Detection of Malicious PDF Files Based on Hierarchical Document Structure (2013, methodology): “Given the increasing interest to the application of learning methods for security problems, some previous work has addressed the methodology for security analysis of learning algorithms [3, 18, 2].”
Unsupervised Learning Trojan (2014, methodology): “The model also contemplates the adversary’s target system and the adversary’s capabilities in similar organizational manner as that of Laskov & Kloft (2009).”