TESSERACT: Eliminating Experimental Bias in Malware Classification across Space and Time

By Feargus Pendlebury, Fabio Pierazzi, Roberto Jordaney, Johannes Kinder, Lorenzo Cavallaro. USENIX Security Symposium, 2018

474 citations recorded by the source corpus.

Is Android malware classification a solved problem? Published F1 scores of up to 0.99 appear to leave very little room for improvement. In this paper, we argue that results are commonly inflated due to two pervasive sources of experimental bias: "spatial bias" caused by distributions of training and testing data that are not representative of a real-world deployment; and "temporal bias" caused by incorrect time splits of training and testing sets, leading to impossible configurations. We propose a set of space and time constraints for experiment design that eliminates both sources of bias. We introduce a new metric that summarizes the expected robustness of a classifier in a real-world setting, and we present an algorithm to tune its performance. Finally, we demonstrate how this allows us to evaluate mitigation strategies for time decay such as active learning. We have implemented our solutions in TESSERACT, an open source evaluation framework for comparing malware classifiers in a realistic setting. We used TESSERACT to evaluate three Android malware classifiers from the literature on a dataset of 129K applications spanning over three years. Our evaluation confirms that earlier published results are biased, while also revealing counter-intuitive performance and showing that appropriate tuning can lead to significant improvements.

Subfields: Computer Networks and Communications, Software, Signal Processing

View Paper: https://www.semanticscholar.org/paper/7ab5a2b7d2caa8969c06e1f24dfda2a6f5c654f6

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

1. Jun Zhang
2. Surya Nepal
3. Fabio Pierazzi
4. Lorenzo Cavallaro
5. Feargus Pendlebury
6. Yang Xiang

In Software

1. Surya Nepal
2. Fabio Pierazzi
3. Lorenzo Cavallaro
4. Jacopo Cortellazzi
5. Jun Zhang
6. Feargus Pendlebury
7. Yang Xiang

In Signal Processing

1. Wei Luo
2. Surya Nepal
3. Lei Pan
4. Fabio Pierazzi
5. Lorenzo Cavallaro
6. Junyang Qiu
7. Jacopo Cortellazzi
8. Jun Zhang

Citations in the corpus

• Intriguing Properties of Adversarial ML Attacks in the Problem Space (2019, methodology): “Tesseract [55] demonstrated that, in nonstationary contexts such as Android malware, if time-aware splits are not considered, then the results may be inflated due to concept drift (i.”
• A Survey of Android Malware Detection with Deep Neural Models (2020, methodology): “It is proved in Reference [77] that the published results of some earlier studies were biased over time.”