Adversarial machine learning: evaluation of attack models & defense mechanisms

In recent years, there has been a sharp increase in the use of mobile platforms and particularly devices based on the Android operating system. This rapid use of mobile devices has fueled cybercriminals’ interest in developing and sharing malicious software. Machine learning algorithms can be used t...

Full description

Saved in:

Bibliographic Details
Main Authors:	Perifanis, Vasileios, Tserpes, Iosif, Περηφάνης, Βασίλειος, Τσερπές, Ιωσήφ
Other Authors:	Rizomiliotis, Panagiotis
Language:	en_US
Published:	2020
Subjects:	android malware detection machine learning neural networks adversarial examples ανίχνευση κακόβουλου λογισμικού μηχανική μάθηση νευρωνικά δίκτυα Machine learning Malware (Computer software) Neural networks (Computer science) Android (Electronic resource) Computer security
Online Access:	http://hdl.handle.net/11610/20142
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1828461732329160704
author	Perifanis, Vasileios Tserpes, Iosif Περηφάνης, Βασίλειος Τσερπές, Ιωσήφ
author2	Rizomiliotis, Panagiotis
author_facet	Rizomiliotis, Panagiotis Perifanis, Vasileios Tserpes, Iosif Περηφάνης, Βασίλειος Τσερπές, Ιωσήφ
author_sort	Perifanis, Vasileios
collection	DSpace
description	In recent years, there has been a sharp increase in the use of mobile platforms and particularly devices based on the Android operating system. This rapid use of mobile devices has fueled cybercriminals’ interest in developing and sharing malicious software. Machine learning algorithms can be used to detect malware with extremely high performance. However, many of these algorithms, and mainly neural network models, are vulnerable to changes in the input data, known as adversarial examples, capable of leading a model to produce misclassifications. This weakness is one of the major problems that the research community is called upon to solve. This thesis presents the evolution of malicious software for Android-based devices over time and refers to the extraction of an application’s features to detect malicious activity. In addition, ways of detecting malware through machine learning models are being developed, as well as ways in which an attacker can deceive these models. This work focuses on the experimental demonstration of the efficiency of machine learning models for malware detection and the weakness of these models against small changes in the input data. Finally, methods for defending models are being evaluated and special features of adversarial examples are being discussed.
id	oai:hellanicus.lib.aegean.gr:11610-20142
institution	Hellanicus
language	en_US
publishDate	2020
record_format	dspace
spelling	oai:hellanicus.lib.aegean.gr:11610-201422025-03-17T11:08:47Z Adversarial machine learning: evaluation of attack models & defense mechanisms Κακόβουλη μηχανική μάθηση: αξιολόγηση μοντέλων επίθεσης και μηχανισμών άμυνας Perifanis, Vasileios Tserpes, Iosif Περηφάνης, Βασίλειος Τσερπές, Ιωσήφ Rizomiliotis, Panagiotis Ριζομυλιώτης, Παναγιώτης other android malware detection machine learning neural networks adversarial examples ανίχνευση κακόβουλου λογισμικού μηχανική μάθηση νευρωνικά δίκτυα Machine learning Malware (Computer software) Neural networks (Computer science) Android (Electronic resource) Computer security In recent years, there has been a sharp increase in the use of mobile platforms and particularly devices based on the Android operating system. This rapid use of mobile devices has fueled cybercriminals’ interest in developing and sharing malicious software. Machine learning algorithms can be used to detect malware with extremely high performance. However, many of these algorithms, and mainly neural network models, are vulnerable to changes in the input data, known as adversarial examples, capable of leading a model to produce misclassifications. This weakness is one of the major problems that the research community is called upon to solve. This thesis presents the evolution of malicious software for Android-based devices over time and refers to the extraction of an application’s features to detect malicious activity. In addition, ways of detecting malware through machine learning models are being developed, as well as ways in which an attacker can deceive these models. This work focuses on the experimental demonstration of the efficiency of machine learning models for malware detection and the weakness of these models against small changes in the input data. Finally, methods for defending models are being evaluated and special features of adversarial examples are being discussed. Τα τελευταία χρόνια παρατηρείται ραγδαία αύξηση στην χρήση κινητών πλατφορμών και ιδιαίτερα σε συσκευές που βασίζονται στο λογισμικό σύστημα Android. Η ραγδαία αυτή χρήση των κινητών συσκευών έχει κεντρίσει το ενδιαφέρον κυβερνοεγκληματιών για την ανάπτυξη και διαμοιρασμό κακόβουλου λογισμικού. Οι αλγόριθμοι μηχανικής μάθησης μπορούν να χρησιμοποιηθούν για τον εντοπισμό κακόβουλου λογισμικού, έχοντας εξαιρετικά υψηλές αποδόσεις. Ωστόσο, πολλοί από αυτούς τους αλγορίθμους και ειδικά τα μοντέλα νευρωνικών δικτύων είναι ευάλωτα σε αλλαγές στα δεδομένα εισόδου, γνωστά ως κακόβουλα παραδείγματα, ικανές να οδηγήσουν ένα μοντέλο στην παραγωγή εσφαλμένων ταξινομήσεων. Η αδυναμία αυτή αποτελεί ένα από τα σημαντικότερα προβλήματα που καλείται η ερευνητική κοινότητα να επιλύσει. Η παρούσα διπλωματική εργασία παρουσιάζει την εξέλιξη του κακόβουλου λογισμικού για κινητές συσκευές βασισμένες στο λογισμικό Android με το πέρασμα του χρόνου και γίνεται αναφορά στους τρόπους εξαγωγής χαρακτηριστικών των εφαρμογών με σκοπό την ανίχνευση κακόβουλης δραστηριότητας. Επιπλέον, αναπτύσσονται οι τρόποι ανίχνευσης κακόβουλου λογισμικού μέσω μοντέλων μηχανικής μάθησης, καθώς και οι τρόποι με τους οποίους ένας επιτιθέμενος μπορεί να εξαπατήσει τα μοντέλα αυτά. Η εργασία επικεντρώνεται στην πειραματική απόδειξη της ακρίβειας των μοντέλων μηχανικής μάθησης για τον εντοπισμό κακόβουλου λογισμικού και την αδυναμία των αλγορίθμων έναντι μικρών αλλαγών στα δεδομένα εισόδου. Τέλος, αξιολογούνται μέθοδοι προστασίας των μοντέλων, καθώς και συζητούνται ενδιαφέρουσες ιδιότητες των κακόβουλων παραδειγμάτων. 2020-03-26T13:24:16Z 2020-03-26T13:24:16Z 2020-02 http://hdl.handle.net/11610/20142 en_US Default License 131 σ. application/pdf Σάμος
spellingShingle	android malware detection machine learning neural networks adversarial examples ανίχνευση κακόβουλου λογισμικού μηχανική μάθηση νευρωνικά δίκτυα Machine learning Malware (Computer software) Neural networks (Computer science) Android (Electronic resource) Computer security Perifanis, Vasileios Tserpes, Iosif Περηφάνης, Βασίλειος Τσερπές, Ιωσήφ Adversarial machine learning: evaluation of attack models & defense mechanisms
title	Adversarial machine learning: evaluation of attack models & defense mechanisms
title_full	Adversarial machine learning: evaluation of attack models & defense mechanisms
title_fullStr	Adversarial machine learning: evaluation of attack models & defense mechanisms
title_full_unstemmed	Adversarial machine learning: evaluation of attack models & defense mechanisms
title_short	Adversarial machine learning: evaluation of attack models & defense mechanisms
title_sort	adversarial machine learning evaluation of attack models defense mechanisms
topic	android malware detection machine learning neural networks adversarial examples ανίχνευση κακόβουλου λογισμικού μηχανική μάθηση νευρωνικά δίκτυα Machine learning Malware (Computer software) Neural networks (Computer science) Android (Electronic resource) Computer security
url	http://hdl.handle.net/11610/20142
work_keys_str_mv	AT perifanisvasileios adversarialmachinelearningevaluationofattackmodelsdefensemechanisms AT tserpesiosif adversarialmachinelearningevaluationofattackmodelsdefensemechanisms AT perēphanēsbasileios adversarialmachinelearningevaluationofattackmodelsdefensemechanisms AT tserpesiōsēph adversarialmachinelearningevaluationofattackmodelsdefensemechanisms AT perifanisvasileios kakoboulēmēchanikēmathēsēaxiologēsēmontelōnepithesēskaimēchanismōnamynas AT tserpesiosif kakoboulēmēchanikēmathēsēaxiologēsēmontelōnepithesēskaimēchanismōnamynas AT perēphanēsbasileios kakoboulēmēchanikēmathēsēaxiologēsēmontelōnepithesēskaimēchanismōnamynas AT tserpesiōsēph kakoboulēmēchanikēmathēsēaxiologēsēmontelōnepithesēskaimēchanismōnamynas

Adversarial machine learning: evaluation of attack models & defense mechanisms

Similar Items