University of Ioannina
Algorithmic techniques for detection and classification of digital objects
Abstract
dc:descriptionIn this PhD Thesis there have been studied the algorithmic techniques for the detection and classification of digital object. In the area of digital objects, this Thesis focuses mainly on the investigation of designing and proposing algorithmic techniques that detect and classify (in terms of indexing) a specific category of digital objects, the one of software, and more precisely the malicious software, providing finally an integrated algorithmic framework for protection against malicious software. Malicious authors, in order to avoid traditional detection methods, have developed highly sophisticated practices focusing on mutating their produced malicious samples, incorporating mutation engines that mutate the structure of the generated malicious samples (i.e., polymorphism and metamorphism). On the first development axis of this thesis, the research focuses on the design and the proposal of a mutation-tolerant graph-based representation of malicious software sample's behavior (behavioral graph) resulted from System-call Dependency Graphs, or, for short ScDG, a Directed Acyclic Graph produced through Dynamic Taint Analysis of the executed sample. So, in the first state we propose the Group Relation Graph, or, for short GrG, a Directed Weighted Graph that is an abstraction of ScDG resulting after grouping disjoint vertices of it, utilizing the property that system-calls can be merged into groups based on their similar functionality. Further, we extent this approach by proposing the Coverage Graph, or, for short CvG, where we investigating the dominating relations among the vertices of GrGs regarding the vertex weight and degree. Additionally, extending the potentials of the above graph-based representations, we also propose the Temporal Graphs, that actually depict the structural evolution of the previously proposed graphs (i.e., GrG and CvG) by depicting their structures through instances captured over specific periods. Among others, we propose a set of similarity metrics that utilize quantitative, relational and qualitative characteristics of the above graph-based representations of malicious software sample's behavior, utilizing them in order to experimentally evaluate the detection and classification potentials of our model. Moreover, since the usage of mobile devices exhibits a wide spread, dependently of te adoption of IoT, throughout this thesis, it has also been studied the development of graph-based algorithmic techniques that would integrate the overall algorithmic framework for protection against malicious software by investigating graph-based strategies for suppressing and finally avoiding potential pandemics caused by malware's spread. More precisely, we propose a set of graph-based techniques for modeling the topology of towns-planning, the node mobility patterns as also the propagation behavior, incorporating them to develop an algorithmic technique that defines the maximum permitted time required by a counter- measure to take effect removing the malware from an infected device (i.e., response time) in order to finally the pandemic spread. Finally, the precision of the proposed approach is tested throughout a series of repetitive (Monte Carlo) series of experiments of various epidemic models and set of factors that affect the malware's spread.
Degree
thesis:*- Grantor dc:publisher
- University of Ioannina
- Year dc:date
- 2019
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Polenakis, Iosif
- Πολενάκης, Ιωσήφ
Subjects
dc:subject × 24- Ασφάλεια πληροφοριακών συστημάτων
- Αλγόριθμοι
- Γραφήματα
- Κακόβουλο λογισμικό
- Ανάλυση κακόβουλου λογισμικού
- Ανίχνευση κακόβουλου λογισμικού
- Κατάταξη κακόβουλου λογισμικού
- Επιδημιολογία
- Αποφυγή πανδημιών
- Information systems' security
- Algorithms
- Graphs
- Malicious software
- Malicious software analysis
- Malicious software detection
- Malicious software classification
- Epidemiology
- Pandemic prevention
- Φυσικές Επιστήμες
- Μαθηματικά
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Mathematics
- Computer and Information Sciences
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/46311