Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Engage Grants Program

Université Laval

$25,000.00

Machine Learning for Android Malware Analysis

EF 2014-15

Résumé
Thales Research & Technology Canada, based in Quebec City, wants to understand how malware infectsAndroid devices in order to protect its multiple Android-based products. For this purpose, Thales has amalware analysis sandbox (MAS), which is a secured environment to observe malware while it is running. ThisMAS produces reports that contains observed actions (API calls, network communications, etc.) performed bythe application under different stimuli exerted by the MAS. Furthermore, following its analysis, the MAS willtell if a sample is a malware or a legitimate application. However, the false positive rate of the MAS'predictions is too large (>50%). The company-specific problem is to find a predictor which has a lower falsepositive rate while not having a larger false negative rate than the MAS' predictions. Since we have at ourdisposal a database of examples of applications which have been classified by experts has being eitherlegitimate or malware, we can view Android malware detection as a classification problem: given this set ofexamples the goal is to produce a classifier that will predict as accurately as possible if a new application ismalware or not based on the application's behaviour in the MAS. To tackle this problem we plan to use threedifferent learning algorithms that we are experienced with and that perform well under different circumstances.The set covering machine (SCM) needs a set of binary-valued features for which there exists a smallconjunction/disjunction of them which is accurate on the training data. AdaBoost needs a set of de-correlatedfeatures for which there exists an accurate majority vote on the training data. Finally the support vectormachine (SVM) will perform well if the exists an hyperplane having a large soft-margin on the training data.The challenge is to find the appropriate feature sets (for AdaBoost and the SCM) and kernels (for the SVM)constructed from the MAS reports that will enable these learning algorithms to deliver a classifier having goodgeneralization performance (i.e., low error rates).
Programme
Engage Grants Program
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
470080-2014
Name nom
Marchand, Mario
Department département
Informatique et de génie logiciel
Organization id
51
Institution établissement
Université Laval
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2014
Competition year année de concours
2014
Award amount
25000
Program id
EGP
Program name en
Engage Grants Program
Program name fr
Programme de subventions d'engagement partenariat
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
1552
Committee name en
Quebec Internal Decision Committee
Committee name fr
Comité de décision interne pour le Québec
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2805
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Learning and inference theories
Research subject fr
Théories de l'apprentissage et l'inférence
Partie
1
Nb partie
1
Application title
Machine Learning for Android Malware Analysis
Keyword
NSERC
Application summary
Thales Research & Technology Canada, based in Quebec City, wants to understand how malware infectsAndroid devices in order to protect its multiple Android-based products. For this purpose, Thales has amalware analysis sandbox (MAS), which is a secured environment to observe malware while it is running. ThisMAS produces reports that contains observed actions (API calls, network communications, etc.) performed bythe application under different stimuli exerted by the MAS. Furthermore, following its analysis, the MAS willtell if a sample is a malware or a legitimate application. However, the false positive rate of the MAS'predictions is too large (>50%). The company-specific problem is to find a predictor which has a lower falsepositive rate while not having a larger false negative rate than the MAS' predictions. Since we have at ourdisposal a database of examples of applications which have been classified by experts has being eitherlegitimate or malware, we can view Android malware detection as a classification problem: given this set ofexamples the goal is to produce a classifier that will predict as accurately as possible if a new application ismalware or not based on the application's behaviour in the MAS. To tackle this problem we plan to use threedifferent learning algorithms that we are experienced with and that perform well under different circumstances.The set covering machine (SCM) needs a set of binary-valued features for which there exists a smallconjunction/disjunction of them which is accurate on the training data. AdaBoost needs a set of de-correlatedfeatures for which there exists an accurate majority vote on the training data. Finally the support vectormachine (SVM) will perform well if the exists an hyperplane having a large soft-margin on the training data.The challenge is to find the appropriate feature sets (for AdaBoost and the SCM) and kernels (for the SVM)constructed from the MAS reports that will enable these learning algorithms to deliver a classifier having goodgeneralization performance (i.e., low error rates).