Poisoning behavioral malware clustering
By Battista Biggio, Konrad Rieck, Davide Ariu, Christian Wressnegger, Igino Corona, Giorgio Giacinto, Fabio Roli. 2014
157 citations recorded by the source corpus.
Clustering algorithms have become a popular tool in computer security to analyze the behavior of malware variants, identify novel malware families, and generate signatures for antivirus systems. However, the suitability of clustering algorithms for security-sensitive settings has been recently questioned by showing that they can be significantly compromised if an attacker can exercise some control over the input data. In this paper, we revisit this problem by focusing on behavioral malware clustering approaches, and investigate whether and to what extent an attacker may be able to subvert these approaches through a careful injection of samples with poisoning behavior. To this end, we present a case study on Malheur, an open-source tool for behavioral malware clustering. Our experiments not only demonstrate that this tool is vulnerable to poisoning attacks, but also that it can be significantly compromised even if the attacker can only inject a very small percentage of attacks into the input data. As a remedy, we discuss possible countermeasures and highlight the need for more secure clustering algorithms.
Subfields: Computer Networks and Communications, Information Systems, Safety, Robustness & Interpretability, Anomaly & Outlier Detection, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Networks and Communications
- Battista Biggio
- Fabio Roli
- Giorgio Giacinto
- Shishir Nagaraja
- Joseph Gardiner
- Jacob Steinhardt
- Nicholas Carlini
- Percy Liang
In Information Systems
- Fabio Roli
- Battista Biggio
- Joseph Gardiner
- Shishir Nagaraja
- Giorgio Giacinto
- Nicolas Papernot
- Percy Liang
- Nicholas Carlini
In Safety, Robustness & Interpretability
- Nicolas Papernot
- Nicholas Carlini
- Patrick Drew McDaniel
- Battista Biggio
- Percy Liang
- Jacob Steinhardt
- Bo Li
- Fabio Roli
In Anomaly & Outlier Detection
In Signal Processing
- Battista Biggio
- Fabio Roli
- Nicholas Carlini
- Nicolas Papernot
- Giorgio Giacinto
- Patrick Drew McDaniel
- Ambra Demontis
- Shishir Nagaraja
Citations in the corpus
- Distillation as a Defense to Adversarial Perturbations Against Deep Neural Networks (2015, background): “For instance, crafting adversarial samples from real malware to evade existing detection methods will require different metrics and perturbations [16], [37].”
- Certified Defenses for Data Poisoning Attacks (2017, background): “There have also been a number of demonstrated vulnerabilities in deployed systems (Newsome et al., 2006; Laskov and Šrndi`c, 2014; Biggio et al., 2014b).”
- Towards Poisoning of Deep Learning Algorithms with Back-gradient Optimization (2017, background): “to cause specificor generic errors. Here we do not follow their naming convention, as it can cause confusion with the interpretation of targeted and indiscriminate attacks introduced in previous work [1, 2, 4, 7, 9, 10, 16, 39]. attacker knows the learning algorithm, but she is not able to derive an”
- The Limitations of Deep Learning in Adversarial Settings (2015, background): “We believe this makes adversarial crafting much easier for input domains like malware executables, which are not as easy to perturb as images [10], [15].”
- Certified Defenses against Adversarial Examples (2018, background): “Newsome et al. (2006), Laskov & Šrndi ` c (2014), and Biggio et al. (2014b) have demonstrated poisoning attacks against real-world systems.”
- SoK: Security and Privacy in Machine Learning (2018, methodology): “The approach was later revisited in the context of malware clustering [41]: malware is modified to include additional behavioral features that place it between two existing clusters in the model’s input domain, thus eventually decreasing the distance between clusters.”
- Attacks on Machine Learning: Lurking Danger for Accountability (2019, methodology): “Complete-linkage Hierarchical Clustering Unsupervised No Poisoning Attack [9] Single-Linkage Hierarchical Clustering Unsupervised No Poisoning Attack [13] Obfuscation Attack [13, 14] Decision Tree/Random Forest Supervised Yes/No Poisoning Attack [46] No Path-finding Attack [72] Model Inversion [26] ”
- A Comprehensive Survey on Poisoning Attacks and Countermeasures in Machine Learning (2022, background): “They can be applied to both supervised learning and unsupervised learning [5, 9, 12, 106].”
- On the Security of Machine Learning in Malware C&C Detection (2016, methodology): “Defenders can take both reactive and proactive approaches [Biggio et al. 2014a].”
- Poisoning the Unlabeled Dataset of Semi-Supervised Learning (2021, methodology): “This essentially connects the sample x′ to the sample x∗, similar to the “bridge” examples from unsupervised clustering attacks [7].”
- Practical Attacks Against Graph-based Clustering (2017, background): “[12, 13] are the first to study adversarial clustering.”
- A Survey on Security Threats and Defensive Techniques of Machine Learning: A Data Driven View (2018, background): “However, some also introduced how to launch the poisoning attack against single-linkage and completelinkage hierarchical clustering [25], [26], [28].”
- Analysis of Causative Attacks against SVMs Learning from Data Streams (2017, methodology): “Critical problems have been tackled successfully via the use of these algorithms, including spam filtering, network anomaly [8] and malware detection [7], and even activity learning tasks in a pervasive setting[9, 14].”
- Novel Feature Extraction, Selection and Fusion for Effective Malware Family Classification (2015, methodology): “The proposed method has not yet been tested for robustness against evasion attacks [11, 28] or poisoning attacks [13, 12] because these kinds of attacks are more frequent against malware detectors rather than against malware classifiers.”
- Exploring Backdoor Poisoning Attacks Against Malware Classifiers (2020, unknown): “Successive work [14], demonstrated the relevance of poisoning attacks in computer security by attacking the Malheur malware behavioral clustering tool [40].”