Picking on the family: Disrupting android malware triage by forcing misclassification

Calleja, Alejandro; Martín, Alfonso Jiménez; Menéndez, Héctor D.; Tapiador, Juan E.; Clark, David

doi:10.1016/j.eswa.2017.11.032

Cited by 49 publications

(43 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Given that manipulating the content of Android malware can be relatively easy, especially due to the possibility of injecting dead code, this behavior highlights the potential vulnerability of such classifiers. In fact, if the decisions of a classifier rely on few features, it is intuitive that detection can be easily evaded by manipulating only few of them, as also confirmed in previous work [6], [8]. Conversely, if a model distributes relevance more evenly among features, evasion may be more difficult (i.e., require manipulating a higher number of features, which may not be always feasible).…”

Section: B Global Explanationsmentioning

confidence: 62%

Explaining Black-box Android Malware Detection

Melis

Maiorca

Biggio

et al. 2018

2018 26th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

Machine-learning models have been recently used for detecting malicious Android applications, reporting impressive performances on benchmark datasets, even when trained only on features statically extracted from the application, such as system calls and permissions. However, recent findings have highlighted the fragility of such in-vitro evaluations with benchmark datasets, showing that very few changes to the content of Android malware may suffice to evade detection. How can we thus trust that a malware detector performing well on benchmark data will continue to do so when deployed in an operating environment? To mitigate this issue, the most popular Android malware detectors use linear, explainable machine-learning models to easily identify the most influential features contributing to each decision. In this work, we generalize this approach to any black-box machinelearning model, by leveraging a gradient-based approach to identify the most influential local features. This enables using nonlinear models to potentially increase accuracy without sacrificing interpretability of decisions. Our approach also highlights the global characteristics learned by the model to discriminate between benign and malware applications. Finally, as shown by our empirical analysis on a popular Android malware detection task, it also helps identifying potential vulnerabilities of linear and nonlinear models against adversarial manipulations.

show abstract

Section: B Global Explanationsmentioning

confidence: 62%

Explaining Black-box Android Malware Detection

Melis

Maiorca

Biggio

et al. 2018

2018 26th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

show abstract

“…The work of Biggio and others [ 58 ] has shown several examples on how these techniques can be easily defeated. Besides, in the malware detection field, tools such as EvadeML [ 59 ], IagoDroid [ 5 ], or EEE [ 9 ] have proved that those features representing the malware can help to learn how to create undetectable variants.…”

Section: Related Workmentioning

confidence: 99%

“…Nevertheless, different works have strong caveats on static analysis, providing malicious concealment via metamorphism or polymorphism, and dynamic analysis via red pills. Besides, modern methods can even attack the triage process, shaping the malicious piece to look as benign as possible using machine learning [ 5 ].…”

Section: Introductionmentioning

confidence: 99%

Mimicking Anti-Viruses with Machine Learning and Entropy Profiles

Menéndez

Llorente

2019

Entropy

Self Cite

View full text Add to dashboard Cite

The quality of anti-virus software relies on simple patterns extracted from binary files. Although these patterns have proven to work on detecting the specifics of software, they are extremely sensitive to concealment strategies, such as polymorphism or metamorphism. These limitations also make anti-virus software predictable, creating a security breach. Any black hat with enough information about the anti-virus behaviour can make its own copy of the software, without any access to the original implementation or database. In this work, we show how this is indeed possible by combining entropy patterns with classification algorithms. Our results, applied to 57 different anti-virus engines, show that we can mimic their behaviour with an accuracy close to 98% in the best case and 75% in the worst, applied on Windows’ disk resident malware.

show abstract

“…SigMal [27] shares some similarities with Bitshred, but uses signal processing based analysis to improve resistance to noise. More recently, Calleja et al [18] showed that confusing statistical classification systems may be easy for malware writers. More generally, the recent research trend on adversarial machine learning [24] cast a shadow on the robustness of triage solutions based on statistical models.…”

Section: Related Workmentioning

confidence: 99%

Triage of IoT Attacks Through Process Mining

Coltellese

Maggi

Marrella

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

The impressive growth of the IoT we witnessed in the recent years came together with a surge in cyber attacks that target it. Factories adhering to digital transformation programs are quickly adopting the IoT paradigm and are thus increasingly exposed to a large number of cyber threats that need to be detected, analyzed and appropriately mitigated. In this scenario, a common approach that is used in large organizations is to setup an attack triage system where security operators, supported by appropriate tools, can cherrypick new attack patterns, which require further in-depth investigation, from a mass of known attacks, which can be managed by automatic means. In this paper, we propose an attack triage system that helps operators quickly identify attacks with unknown behaviors, and later analyze them in detail. The novelty introduced by our solution is in the usage of process mining techniques to model known attacks and identify new variants. We demonstrate the feasibility of our approach through a twofold evaluation based on three well-known IoT botnets, BASHLITE, LIGHTAIDRA and MIRAI, and on real current attack patterns collected through an IoT honeypot.

show abstract

Picking on the family: Disrupting android malware triage by forcing misclassification

Cited by 49 publications

References 10 publications

Explaining Black-box Android Malware Detection

Explaining Black-box Android Malware Detection

Mimicking Anti-Viruses with Machine Learning and Entropy Profiles

Triage of IoT Attacks Through Process Mining

Contact Info

Product

Resources

About