Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Omwille van de steeds vernieuwende ontwikkelingen op het gebied van draadloze communicatie en sensor technologie hebben sinds de jaren '90 contextgedreven toepassingen meer en meer aandacht gekregen en dit heeft geleid tot nieuwe computing paradigma's die luisteren naar de namen Ubiquitous and Pervasive Computing en Ambient Intelligence. Nieuwe applicaties zijn op de hoogte van de context van hun gebruiker en passen zich hieraan automatisch aan. Men zegt dat dergelijke applicaties contextbewust zijn. Hiermee bedoelt men dat dergelijke applicaties relevante informatie over de situatie van de gebruiker en zijn omgeving gebruiken, zoals zijn huidige locatie en tijdstip, maar ook de activiteiten die op dat ogenblik uitgevoerd worden en gebruikersvoorkeuren die hier aan gekoppeld zijn.De ontwikkeling van contextbewuste applicaties die draaien in zeer dynamische omgevingen en die zich aanpassen aan de veranderende behoeften van de gebruikers, blijft een complexe en foutgevoelige taak. Om deze situatie te verhelpen, is dit proefschrift gericht op het verstrekken van middleware ondersteuning voor de ontwikkeling van contextbewuste applicaties in een Ambient Intelligence (AmI) omgeving. De voornaamste doelstellingen zijn het verschaffen van modelleerabstracties voor contextinformatie en toepassingen, en middleware ondersteuning voor het behandelen van contextinformatie en het aanpassen van applicaties aan deze context. Door het beheer van contextinformatie te delegeren naar de middleware verminderen we de ontwikkeltijd die nodig is om contextbewust gedrag door de applicatie te realiseren.Het proefschrift bevat verscheidene onderzoeksbijdragen. Ten eerste, omdat contextbewuste toepassingen informatie nodig hebben om hun gedrag te bepalen en om te beslissen met wie samen te werken, presenteert deze thesis abstracties om context te modelleren zodat elke applicatie deze op dezelfde manier interpreteert. Deze abstracties omvatten semantische, ruimtelijke, temporele aspecten en eveneens manieren om contextambiguiteit uit te sluiten. Ten tweede voorzien we middleware-ondersteuning die gemeenschappelijke functionaliteit implementeert voor het beheren en verwerken van contextinformatie. Het encapsuleert deze functies als modulaire bouwstenen die kunnen worden geactiveerd wanneer deze nodig zijn. Deze bouwstenen zijn verantwoordelijk voor de verwerving van de contextinformatie uit verschillende bronnen, de verwerking en de persistentie ervan. Ten derde leveren we ondersteuning voor contextgedreven aanpassing van mobiele diensten gebaseerd op modelleerabstracties die zowel functionele als niet-functionele aspecten van contextbewuste applicaties beschrijven. Ten vierde presenteert dit proefschrift algoritmen en protocollen om contextinformatie te distribueren van en naar relevante toepassingen die actief zijn in mobiele ad-hoc netwerken, evenals mechanismen voor de contextgedreven ontdekking, selectie en het gebruik van externe applicaties. Door het delen van contextinformatie in het netwerk kunnen externe toepassingen profiteren van de context die zij zelf niet kunnen waarnemen.Het proefschrift presenteert real-world toepassingen en case studies die de implementatie van de middleware en de onderliggende algoritmen en mechanismen evalueren, en die het proces en de problemen in verband met het ontwerpen van contextbewuste toepassingen illustreren.

681.3*C3 <043> --- 681.3*C24 <043> --- 681.3*H4 <043> --- Academic collection --- 681.3*C24 <043> Distributed systems: distributed databases; distributed applications; networkoperating systems--Dissertaties --- Distributed systems: distributed databases; distributed applications; networkoperating systems--Dissertaties --- 681.3*C3 <043> Special-purpose and application-based systems: microprocessor/microcomputer; process control-, real-time, signal processing systems (Computer systems organization)--See also {681.3*J7}--Dissertaties --- Special-purpose and application-based systems: microprocessor/microcomputer; process control-, real-time, signal processing systems (Computer systems organization)--See also {681.3*J7}--Dissertaties --- Computer science--Information systems applications (GIS etc.)--Dissertaties --- Theses

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

ISAmI is the International Symposium on Ambient Intelligence, and aims to bring together researchers from various disciplines that are interested in all aspects of Ambient Intelligence. The symposium provides a forum to present and discuss the latest results, innovative projects, new ideas and research directions, and to review current trends in this area. This volume presents the papers that have been accepted for the 2011 edition, both for the main event and workshop. The ISAmI workshop WoRIE promises to be a very interesting event that complements the regular program with an emerging topic on reliability of intelligent environments

Artificial intelligence. Robotics. Simulation. Graphics --- neuronale netwerken --- fuzzy logic --- cybernetica --- KI (kunstmatige intelligentie) --- robots

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Today's digital society is threatened by evermore sophisticated cyber attacks. One of the drivers behind this phenomenon is the advances in Artificial Intelligence (AI), which empower attackers to perform automated analysis of victim systems. Thankfully, AI can also be leveraged to upgrade defensive systems to match the unprecedented speed and scale of attacks. This duality of use motivates comprehensive research into and scientific understanding of both offensive and defensive applications of AI in cybersecurity. In this dissertation, we investigate deep learning as a compelling data-driven AI technique in the realm of network security and privacy. These domains deal with network traffic characterized by overwhelming volumes, dynamic nature, and low interpretability by human analysts. Hence, we focus on applying deep learning techniques to automatically derive high-level representations from network traffic for offensive and defensive approaches. The overarching objective is then to assess general applicability, advantages, and possible drawbacks of deep neural networks applied to security and privacy. In the first part of this dissertation, we research side-channel traffic analysis attacks on Tor, a popular privacy-enhancing technology on the web. As a first contribution, we develop deep learning techniques to deanonymize encrypted Tor traffic automatically. Our results, for the first time, obviated the need for manually engineering traffic features, as required by prior website fingerprinting attacks. This work has proven to be a game changer in this area, leading to further active development of deep learning-based website fingerprinting attacks and more robust defenses. As a second contribution, we propose a more realistic data collection setup to diminish timing bias when modeling so-called end-to-end traffic confirmation attacks on Tor in a lab environment. These results show that deep neural networks, now the established state-of-the-art approach for traffic analysis on Tor, silently memorize artifacts in data, which may cause a drastic overestimation of attack performance. We develop a general scientific methodology for replicating and comparing attacks while considering the non-determinism and variability of deep neural networks. In the second part of this dissertation, we investigate the promises and pitfalls of deep learning for defensive applications in network intrusion detection. First, we comprehensively overview the foundations, recent advances, and open challenges for data-driven network intrusion detection. Next, we present an empirical analysis of the applicability and reliability of deep neural networks as a stand-alone mechanism for analyzing raw traffic and detecting malicious or suspicious behaviors. These results highlight some of the inherent critical drawbacks of applying deep neural networks to real-world defensive applications.In summary, our contributions and resulting insights provide a solid ground for further scientific advances in data-driven security and privacy research while promoting a holistic view of the decision-making capacities of deep learning.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Automatic border control systems have recently been shown to be vulnerable to a morphing attack. It attacks the authentication process by creating face images that can be used to match multiple people. This can then be used to create a passport that is valid for two or more individuals. Current methods to generate these morphs have been shown to have several problems: They are (1) very easy to identify using specialized detectors and (2) only work when there is a high similarity between the two input faces. Our goal is to find a new generation method that does not suffer from these shortcomings. To do this we explore generating morphs using generative adversarial networks. First we create an overview of the conditional neural network designs used for image generation. For each we discuss the core idea behind it and its advantages and disadvantages. We then work out two of these designs for the application of morph generation. Finally we compare their performance to the current state of the art of morph generation. To do so we introduce a new evaluation methodology that tests both the quality and morph similarity with human and automatic evaluation methods. Our approach achieves the same automated match rate of 60% compared to existing techniques, but results in morphs that are visually a better split between the inputs. The downside however is that our method has more artifacting. We conclude that generative adversarial neural networks show potential for morph generation, but miss a good way to teach the network about the similarity between two faces, which directly results in poorer quality morphs.