Can We Leverage Predictive Uncertainty to Detect Dataset Shift and Adversarial Examples in Android Malware Detection? (2021-09-20T00:00:00.000000Z)

TL;DR

Predictive uncertainty indeed helps achieve reliable malware detection in the presence of dataset shift, but cannot cope with adversarial evasion attacks; approximate Bayesian methods are promising to calibrate and generalize malware detectors to deal with dataset shift; but can not cope with adversary evasion attacks.

Abstract

The deep learning approach to detecting malicious software (malware) is promising but has yet to tackle the problem of dataset shift, namely that the joint distribution of examples and their labels associated with the test set is different from that of the training set. This problem causes the degradation of deep learning models without users’ notice. In order to alleviate the problem, one approach is to let a classifier not only predict the label on a given example but also present its uncertainty (or confidence) on the predicted label, whereby a defender can decide whether to use the predicted label or not. While intuitive and clearly important, the capabilities and limitations of this approach have not been well understood. In this paper, we conduct an empirical study to evaluate the quality of predictive uncertainties of malware detectors. Specifically, we re-design and build 24 Android malware detectors (by transforming four off-the-shelf detectors with six calibration methods) and quantify their uncertainties with nine metrics, including three metrics dealing with data imbalance. Our main findings are: (i) predictive uncertainty indeed helps achieve reliable malware detection in the presence of dataset shift, but cannot cope with adversarial evasion attacks; (ii) approximate Bayesian methods are promising to calibrate and generalize malware detectors to deal with dataset shift, but cannot cope with adversarial evasion attacks; (iii) adversarial evasion attacks can render calibration methods useless, and it is an open problem to quantify the uncertainty associated with the predicted labels of adversarial examples (i.e., it is not effective to use predictive uncertainty to detect adversarial examples).

References51 items

Leveraging Uncertainty for Improved Static Malware Detection Under Extreme False Positive Constraints

Learning Under Concept Drift

Uncertainty Estimation and Calibration with Finite-State Probabilistic RNNs

Uncertainty as a Form of Transparency: Measuring, Communicating, and Using Uncertainty

Enhancing State-of-the-art Classifiers with API Semantics to Detect Evolved Android Malware

Adversarial Deep Ensemble: Evasion Attacks and Defenses for Malware Detection

Learning under Concept Drift: A Review

Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift

A Multimodal Deep Learning Method for Android Malware Detection Using Various Features

Evaluating model calibration in classification

Uncertainty in Neural Networks: Approximately Bayesian Ensembling

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

Statistical Estimation of Malware Detection Metrics in the Absence of Ground Truth

Adversarial Examples for Malware Detection

Adversarial Machine Learning in Malware Detection: Arms Race between Evasion Attack and Defense

A Survey on Malware Detection Using Data Mining Techniques

On Calibration of Modern Neural Networks

R2-D2: ColoR-inspired Convolutional NeuRal Network (CNN)-based AndroiD Malware Detections

Yes, Machine Learning Can Be More Secure! A Case Study on Android Malware Detection

Deep Android Malware Detection

What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?

A Survey on Systems Security Metrics

Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles

Leveraging uncertainty information from deep neural networks for disease detection

AndroZoo: Collecting Millions of Android Apps for the Research Community

Weight Uncertainty in Neural Network

Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

Weight Uncertainty in Neural Networks

Obtaining Well Calibrated Probabilities Using Bayesian Binning

Convolutional Neural Networks for Sentence Classification

An evasion and counter-evasion study in malicious websites detection

Practical Variational Inference for Neural Networks

Bayesian Learning via Stochastic Gradient Langevin Dynamics

The Balanced Accuracy and Its Posterior Distribution

Dataset Shift in Machine Learning

Bidirectional Long Short-Term Memory Networks for Predicting the Subcellular Localization of Eukaryotic Proteins

Predicting good probabilities with supervised learning

Evaluating Predictive Uncertainty Challenge

Learning in the Presence of Concept Drift and Hidden Contexts

Forensics Corvus

Kaspersky Lab

Droidetec: Androidmalwaredetectionandmaliciouscodelocalizationthroughdeeplearning

Transcend: Detecting Concept Drift in Malware Classification Models

An Overview of Concept Drift Applications

TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems

DREBIN: Effective and Explainable Detection of Android Malware in Your Pocket

Dropout: a simple way to prevent neural networks from overfitting

Machine Learning Challenges. Evaluating Predictive Uncertainty, Visual Object Classification, and Recognising Tectual Entailment

VERIFICATION OF FORECASTS EXPRESSED IN TERMS OF PROBABILITY

This Paper Is Included in the Proceedings of the 12th Usenix Symposium on Operating Systems Design and Implementation (osdi '16). Tensorflow: a System for Large-scale Machine Learning Tensorflow: a System for Large-scale Machine Learning

under temporal covariate shifts. We observe that the Accuracy, and BSE are smaller than their balanced counterparts plotted