<p dir="ltr">Deep Neural Networks (DNNs) have been deployed on edge devices for numerous applications, ranging from computer vision, speech recognition, and anomaly detection. When deployed on edge devices, dedicated DNN compilers are used to compile DNNs into binaries to exploit instruction set architectures’ (ISAs’) features and hardware accelerators (e.g., NPU, GPU). These DNN binaries on edge devices process sensitive user information, conduct critical missions, and are considered confidential intellectual property.</p><p dir="ltr">From the security standpoint, the ability to reverse engineer such binaries (i.e., recovering the original, high-level representation of the implemented DNN) enables several applications, such as DNN models stealing, gray/white-box adversarial machine learning attacks and defenses, and backdoor detection. However, no existing reverse engineering technique can recover a high-level representation of a DNN model from its compiled binary code.</p><p dir="ltr">In this dissertation, we propose the following pioneering research for reverse engineering DNN on the edge device. (i) We design and implement the first compiler- and ISA-agnostic DNN decompiler, DnD, with the static analysis technique, capable of extracting DNN models from DNN binaries running on CPU-only devices without the hardware accelerator. We show that our decompiler can perfectly recover DNN models from different DNN binaries. Furthermore, it can extract DNN models used by real-world micro-controllers and enable white-box adversarial machine learning attacks against the DNN models. (ii) We design and implement a novel data-driven approach, NeuroScope, based on dynamic analysis and machine learning to reverse engineer DNN binaries. This compiler-independent and code-feature-free approach supports a larger variety of DNN binaries across different DNN compilers and hardware platforms. We demonstrate its capability by using it to reverse engineer DNN binaries unsupported by previous approaches with high accuracy. Moreover, we showcase how NeuroScope can be used to reverse engineer a proprietary DNN binary compiled with a closed-source compiler and enable gray-box adversarial machine learning attacks.</p>

Android forensics has evolved over time offering significant opportunities and exciting challenges. On one hand, being an open source platform Android is giving developers the freedom to contribute to the rapid growth of the Android market whereas on the other hand Android users may not be aware of the security and privacy implications of installing these applications on their phones. Users may assume that a password-locked device protects their personal information, but applications may retain private information on devices, in ways that users might not anticipate. In this thesis we will be concentrating on one such application called 'WhatsApp', a popular social networking application. We will be forming an outline on how forensic investigators can extract useful information from WhatsApp and from similar applications installed on an Android platform. Our area of focus is extraction and analysis of application user data from non-volatile external storage and the volatile memory (RAM) of an Android device.

The anomaly detection task is to recognize the presence of an unusual (and potentially hazardous) state within the behaviors or activities of a computer user, system, or network with respect to some model of normal behavior which may be either hard-coded or learned from observation. An anomaly detection agent faces many learning problems including learning from streams of temporal data, learning from instances of a single class, and adaptation to a dynamically changing concept. The domain is complicated by considerations of the trusted insider problem (recognizing the difference between innocuous and malicious behavior changes on the part of a trusted user). This study introduces the anomaly detection in web applications and formulates it as a machine learning task on temporal sequence data. In this study the goal is to develop a model or profile of normal working state of web application user and to detect anomalous conditions as deviations from the expected behavior patterns. We focus, here, on learning models of normality at the user behavioral level, as observed through a web application. In this study we introduce some sensors intended to function as a focus of attention unit at the lowest level of a classification hierarchy using Finite State Markov Chains and Hidden Markov Models and discuss the success of these sensors.

Στην παρούσα διπλωματική εργασία παρουσιάζονται βασικές έννοιες και μέθοδοι για την ασφάλεια ιστοσελίδων και ιδιαίτερα των site με web application προσανατολισμό, χωρίς αυτό να σημαίνει ότι αρκετές τεχνικές προστασίας και σφάλματα που θα εντοπίσουμε δεν μπορούν να συναντηθούν και σε άλλου σκοπού ιστοσελίδες. Αρχικά, γίνεται αναφορά στο τι είναι μια εφαρμογή ιστού (web app) και ποια είναι τα στοιχεία που την αποτελούν. Στη συνέχεια, χρησιμοποιώντας έρευνες, παρουσιάζονται κάποιες από τις πιο “δημοφιλείς” επιθέσεις που γίνονται σε ιστοσελίδες και περιγράφεται πιο διεξοδικά ποια αδύνατα σημεία της δομής των ιστοσελίδων εκμεταλλεύονται. Παράλληλα, γίνεται αναφορά στο πως και με ποια εργαλεία μπορούμε να εντοπίσουμε και να κλείσουμε τα κενά ασφαλείας που τυχόν έχει μία εφαρμογή ιστού. Τέλος, παρουσιάζεται η εφαρμογή που αναπτύχθηκε στα πλαίσια της εργασίας με σκοπό να γίνει επίδειξη συγκεκριμένων επιθέσεων και σφαλμάτων που παρατηρούνται στο διαδίκτυο. / In the following pages basic principals and methods are presented in order to secure websites and web applications. I begin by mentioning what is a web application. Moreover, by using statistics and recent researches from various sources i mention the most common web app attack methods and which vulnerabilities can be found in a web app and how to prevent exploiting, something we can accomplish by using various penetration testing tools. Finally, by using a basic web app some web attacks are shown so that it will become more clear how these attacks work.

Made available in DSpace on 2015-04-11T14:03:18Z (GMT). No. of bitstreams: 1 Angelo Eduardo Nunan.pdf: 2892243 bytes, checksum: 5653024cae1270242c7b4f8228cf0d2c (MD5) Previous issue date: 2012-05-14 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Web applications are currently an important environment for access to services available on the Internet. However, the security assurance of these resources has become an elementary task. The structure of dynamic websites composed by a set of objects such as HTML tags, script functions, hyperlinks and advanced features in web browsers may provide numerous resources and interactive services, for instance e-commerce, Internet banking, social networking, blogs, forums, among others. On the other hand, these features helped to increase the potential security risks and attacks, which are the results of malicious codes injection. In this context, Cross-Site Scripting (XSS) is highlighted at the top of the lists of the greatest threats to web applications in recent years. This work presents a method based on supervised machine learning techniques to detect XSS in web pages. A set of features extracted from URL contents and web document are employed in order to discriminate XSS patterns and to successfully classify both malicious and non-malicious pages / As aplicações web atualmente representam um importante ambiente de acesso aos serviços oferecidos na Internet. Garantir a segurança desses recursos se tornou uma tarefa elementar. A estrutura de sites dinâmicos constituída por um conjunto de objetos, tais como tags de HTML, funções de script, hiperlinks e recursos avançados em navegadores web levou a inúmeras funcionalidades e à interatividade de serviços, tais como e-commerce, Internet banking, redes sociais, blogs, fóruns, entre outros. No entanto, esses recursos têm aumentado potencialmente os riscos de segurança e os ataques resultantes da injeção de códigos maliciosos, onde o Cross-Site Scripting aparece em destaque, no topo das listas das maiores ameaças para aplicações web nos últimos anos. Este trabalho apresenta um método baseado em técnicas de aprendizagem de máquina supervisionada para detectar XSS em páginas web, a partir de um conjunto de características extraídas da URL e do documento web, capazes de discriminar padrões de ataques XSS e distinguir páginas web maliciosas das páginas web normais ou benignas

The thesis deals with models used in access control to operating systems' resources. The thesis' goals are the theoretical description of these models and their comparison, the description of their implementations in selected operating systems and the description of their utilization in securing typified servers. In the first chapter, the reader is introduced to basic terms and principles of computer security and access control. In the second, also theoretical part, selected access control models are described from different viewpoints, for example their factual specialization, basic rules, principles and evolution. At the end of the chapter, the models are compared to each other according to specified criteria. The third, more practically oriented chapter, continues from the initial, both general and concrete introduction with operating systems, to description of access control model implementations in selected operating systems. At the end of the chapter, these implementations too are compared according to specified criteria. The fourth, final part, is dedicated to the description of the actual securing of a web and file server using operating system level access control means. The end of the chapter is dedicated to the possibilities of using these means to secure web applications. The contribution of this thesis from the theoretical point of view is a well-arranged and compact access control model comparison, and also the merger of the theoretical base with practical use of the described model implementations. The thesis is recommended to people interested in the computer security issues in general and people interested in access control from both theoretical and practical sides, e.g. system administrators or system designers.

In recent years, companies and organizations have increasingly integrated software security testing into the software development life cycle using DevOps practices. The current integration approach introduces multiple challenges in an information technology environment that consists of a large number of software development projects and multiple software security testing tools. This thesis aims to address these challenges by proposing a microservice-based framework to unify application security testing. The thesis first identifies the challenges, then proposes a design for a framework based on relevant literature and common characteristics of application security testing tools. The main components of the proposed framework are implemented and evaluated. The evaluation result shows that the framework offers many benefits: more secure credential management process, reduced execution time for Continuous Integration (CI) pipelines, and more efficient project onboarding and management. Furthermore, the integration of the proposed framework does not introduce major security threats to the current environment. / Under de senaste åren har företag och organisationer i allt högre grad integrerat testning av programvarusäkerhet i livscykeln för programvaruutveckling med hjälp av DevOps-metoder. Den nuvarande integrationsmetoden medför flera utmaningar i en informationsteknisk miljö som består av ett stort antal programvaruutvecklingsprojekt och flera verktyg för testning av programvarusäkerhet. Detta examensarbete syftar till att ta itu med dessa utmaningar genom att föreslå en mikrotjänstbaserat ramverk för enhetlig testning av programsäkerhet. I arbetet identifieras först utmaningarna och därefter föreslås en konstruktion baserad på relevant litteratur och gemensamma egenskaper hos verktyg för testning av applikationssäkerhet. De viktigaste komponenterna i det föreslagna ramverket implementeras och utvärderas. Utvärderingsresultatet visar att ramverket erbjuder många fördelar: säkrare process för hantering av autentiseringsuppgifter, kortare genomförandetid för Continuous Integration (CI)-pipelines och effektivare projektstart och -hantering. Dessutom medför integrationen av det föreslagna ramverket inga större säkerhetshot i den nuvarande miljön.

The goal of this research is to improve the security of smart home hubs by developing a standard against which hubs can be evaluated. This was done by first reviewing existing standards, guides, and collections of best practices. I determined that adapting or extending an existing standard was the best way to proceed. Potential candidates were selected, and after thorough comparison, I chose to extend the OWASP Application Security Verification Standard (ASVS). Extensions were composed of additional security requirements to address smart home hub functionality not covered by the existing requirements of the ASVS. These additional requirements were developed based upon existing best practices and are referred to as the Smart Home Extensions. Where a best practice or guidance did not yet exist for a particular hub functionality, guidance from related fields was adapted. The entire set of Smart Home Extensions were reviewed by industry experts, updated based on feedback, and then sent on for further peer review. Four smart home hubs – VeraLite, Wink, Connect, and SmartThings – were evaluated using the ASVS with the Smart Home Extensions. The evaluation uncovered security vulnerabilities in all four hubs, some previously disclosed by other researchers, and others new. Analysis of the evaluation data suggests that authentication is a common problem area, among others. Based on the performance of the hubs and the data collected, I suggest that the ASVS and Smart Home Extensions can be an effective tool to provide insight into the security posture of smart home hubs.

<p dir="ltr">In the contemporary digital era, Advanced Persistent Threat (APT) attacks are evolving, becoming increasingly sophisticated, and now perilously targeting critical cyber-physical systems, notably Industrial Control Systems (ICS). The intersection of digital and physical realms in these systems enables APT attacks on ICSs to potentially inflict physical damage, disrupt critical infrastructure, and jeopardize human safety, thereby posing severe consequences for our interconnected world. Provenance tracing techniques are essential for investigating these attacks, yet existing APT attack forensics approaches grapple with scalability and maintainability issues. These approaches often hinge on system- or application-level logging, incurring high space and run-time overheads and potentially encountering difficulties in accessing source code. Their dependency on heuristics and manual rules necessitates perpetual updates by domain-knowledge experts to counteract newly developed attacks. Additionally, while there have been efforts to verify the safety of Programming Logic Controller (PLC) code as adversaries increasingly target industrial environments, these works either exclusively consider PLC program code without connecting to the underlying physical process or only address time-related physical safety issues neglecting other vital physical features.</p><p dir="ltr">This dissertation introduces two novel frameworks, ATLAS and ARCHPLC, to address the aforementioned challenges, offering a synergistic approach to fortifying cybersecurity in the face of evolving APT and ICS threats. ATLAS, an effective and efficient multi-host attack investigation framework, constructs end-to-end APT attack stories from audit logs by combining causality analysis, Natural Language Processing (NLP), and machine learning. Identifying key attack patterns, ATLAS proficiently analyzes and pinpoints attack events, minimizing alert fatigue for cyber analysts. During evaluations involving ten real-world APT attacks executed in a realistic virtual environment, ATLAS demonstrated an ability to recover attack steps and construct attack stories with an average precision of 91.06%, a recall of 97.29%, and an F1-score of 93.76%, providing a robust framework for understanding and mitigating cyber threats.</p><p dir="ltr">Concurrently, ARCHPLC, an advanced approach for enhancing ICS security, combines static analysis of PLC code and data mining from ICS data traces to derive accurate invariants, providing a comprehensive understanding of ICS behavior. ARCHPLC employs physical causality graph analysis techniques to identify cause-effect relationships among plant components (e.g., sensors and actuators), enabling efficient and quantitative discovery of physical causality invariants. Supporting patching and run-time monitoring modes, ARCHPLC inserts derived invariants into PLC code using program synthesis in patching mode and inserts invariants into a dedicated monitoring program for continuous safety checks in run-time monitoring mode. ARCHPLC adeptly detects and mitigates run-time anomalies, providing exceptional protection against cyber-physical attacks with minimal overhead. In evaluations against 11 cyber-physical attacks on a Fischertechnik manufacturing plant and a chemical plant simulator, ARCHPLC protected the plants without any false positives or negatives, with an average run-time overhead of 14.31% in patching mode and 0.4% in run-time monitoring mode.</p><p dir="ltr">In summary, this dissertation provides invaluable solutions that equip cybersecurity professionals to enhance APT attack investigation, enabling them to identify and comprehend complex attacks with heightened accuracy. Moreover, these solutions significantly bolster the safety and security of ICS infrastructure, effectively protecting critical systems and strengthening defenses against cyber-physical attacks, thereby contributing substantially to the field of cybersecurity.</p>

The market of Android applications is huge, and in 2019, Google Play users worldwide downloaded 84.3 billion mobile applications. With such a big user base, any security issues could have big negative impacts. That is why penetration testing of Android applications is important and it is also why Google has a bug bounty program where people can submit vulnerability reports on their most downloaded applications. The aim of the project was to assess the security of Android applications from the Google Play Security Reward Program by performing penetration tests on the applications. A threat model of Android applications was made where potential threats were identified. A choice was made to focus on the Spotify Application for Android where threats were given ratings based on risks associated with them in the context of the Spotify Application. Penetration tests were made where testing depth was determined by the ratings associated with the attacks.The results of the tests showed that the Spotify Application is secure, and no test showed any real possibility of exploiting the application. The perhaps biggest potential exploit found is a Denial of Service attack that can be made through a malicious application interacting with the Spotify application. The result doesn’t guarantee that the application isn’t penetrable and further testing is needed to give the result more reliability. The methods used in the project can however act as a template for further research into both Spotify and other Android applications. / Marknaden för Android applikationer är enorm och 2019 laddade Google Play användare ner 84.3 miljarder mobil-applikationer. Med en så stor användarbas kan potentiella säkerhetsproblem få stora negativa konsekvenser. Det är därför penetrationstest är viktiga och varför Google har ett bug bounty program där folk kan skicka in sårbarhetsrapporter för deras mest nedladdade applikationer. Målet med projektet är att bedöma säkerheten hos Android applikationer från Google Play Security Reward Program genom utförande av penetrationstester på applikationerna. En hotmodell över Android applikationer skapades, där potentiella hot identifierades. Ett val att fokusera på Spotify för Android gjordes, där hot gavs rankingar baserat på riskerna associerade med dem i kontexten av Spotify applikationen. Penetrationstest gjordes med testdjup avgjort av rankingarna associerade med attackerna.Resultatet av testen visade att Spotify applikationen var säker, och inga test visade på några riktiga utnyttjningsmöjligheter av applikationen. Den kanske största utnyttjningsmöjligheten som hittades var en Denial of Service-attack som kunde göras genom en illvillig applikation som interagerar med Spotify applikationen. Resultaten garanterar inte att applikationen inte är penetrerbar och fortsatt testande behövs för att ge resultatet mer trovärdighet. Metoderna som användes i projektet kan i alla fall agera som en mall för fortsatt undersökning av både Spotify såväl som andra Android applikationer.