004 Informatik
Refine
Document Type
- Master's Thesis (21)
- Conference Proceeding (5)
- Bachelor Thesis (3)
- Article (reviewed) (2)
- Doctoral Thesis (2)
Conference Type
- Konferenzartikel (5)
Language
- English (33) (remove)
Keywords
- IT-Sicherheit (6)
- Blockchain (3)
- Deep learning (3)
- Computersicherheit (2)
- Homomorphic Encryption (2)
- Internet der Dinge (2)
- JavaScript (2)
- Künstliche Intelligenz (2)
- Machine Learning (2)
- Maschinelles Lernen (2)
Institute
- Fakultät Medien (M) (ab 22.04.2021) (22)
- Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) (9)
- ivESK - Institut für verlässliche Embedded Systems und Kommunikationselektronik (4)
- Fakultät Medien und Informationswesen (M+I) (bis 21.04.2021) (3)
- Fakultät Wirtschaft (W) (1)
- IMLA - Institute for Machine Learning and Analytics (1)
Open Access
- Closed Access (19)
- Closed (8)
- Open Access (6)
- Diamond (2)
This thesis deals with the implementation of character controls and combat system of the Action Adventure 'Scout 3D'. The game development was realized with the game engine Unity 3D. In the first part, the architecture of a typical game engine is explained. The single components are describes step by step. Then, five well-known game engines are compared and evaluated. In the next chapter, a short overview about design and architecture patterns is worked out. The features of Unity, that are used for the implementation, and Unity's animation system 'Mecanim, are described finally. The second part includes the requirement definitions for the game 'Scout COD' which define player input, different conditions that allow or disallow several activities and the behaviour of enemies. With the help of patterns the architecture of the game is designed. Then, the implementation is explained by means of code snippets.
The recent successes and wide spread application of compute intensive machine learning and data analytics methods have been boosting the usage of the Python programming language on HPC systems. While Python provides many advantages for the users, it has not been designed with a focus on multiuser environments or parallel programming - making it quite challenging to maintain stable and secure Python workflows on a HPC system. In this paper, we analyze the key problems induced by the usage of Python on HPC clusters and sketch appropriate workarounds for efficiently maintaining multi-user Python software environments, securing and restricting resources of Python jobs and containing Python processes, while focusing on Deep Learning applications running on GPU clusters.
The interaction between agents in multiagent-based control systems requires peer to peer communication between agents avoiding central control. The sensor nodes represent agents and produce measurement data every time step. The nodes exchange time series data by using the peer to peer network in order to calculate an aggregation function for solving a problem cooperatively. We investigate the aggregation process of averaging data for time series data of nodes in a peer to peer network by using the grouping algorithm of Cichon et al. 2018. Nodes communicate whether data is new and map data values according to their sizes into a histogram. This map message consists of the subintervals and vectors for estimating the node joining and leaving the subinterval. At each time step, the nodes communicate with each other in synchronous rounds to exchange map messages until the network converges to a common map message. The node calculates the average value of time series data produced by all nodes in the network by using the histogram algorithm. The relative error for comparing the output of averaging time series data, and the ground truth of the average value in the network will decrease as the size of the network increases. We perform simulations which show that the approximate histograms method provides a reasonable approximation of time series data.
Annotated training data is essential for supervised learning methods. Human annotation is costly and laborsome especially if a dataset consists of hundreds of thousands of samples and annotators need to be hired. Crowdsourcing emerged as a solution that makes it easier to get access to large amounts of human annotators. Introducing paid external annotators however introduces malevolent annotations, both intentional and unintentional. Both forms of malevolent annotations have negative effects on further usage of the data and can be summarized as spam. This work explores different approaches to post-hoc detection of spamming users and which kinds of spam can be detected by them. A manual annotation checking process resulted in the creation of a small user spam dataset which is used in this thesis. Finally an outlook for future improvements of these approaches will be made.
The evolution of cellular networks from its first generation (1G) to its fourth generation (4G) was driven by the demand of user-centric downlink capacity also technically called Mobile Broad-Band (MBB). With its fifth generation (5G), Machine Type Communication (MTC) has been added into the target use cases and the upcoming generation of cellular networks is expected to support them. However, such support requires improvements in the existing technologies in terms of latency, reliability, energy efficiency, data rate, scalability, and capacity.
Originally, MTC was designed for low-bandwidth high-latency applications such as, environmental sensing, smart dustbin, etc. Nowadays there is an additional demand around applications with low-latency requirements. Among other well-known challenges for recent cellular networks such as data rate energy efficiency, reliability etc., latency is also not suitable for mission-critical applications such as real-time control of machines, autonomous driving, tactile Internet etc. Therefore, in the currently deployed cellular networks, there is a necessity to reduce the latency and increase the reliability offered by the networks to support use cases such as, cooperative autonomous driving or factory automation, that are grouped under the denomination Ultra-Reliable Low-Latency Communication (URLLC).
This thesis is primarily concerned with the latency into the Universal Terrestrial Radio Access Network (UTRAN) of cellular networks. The overall work is divided into five parts. The first part presents the state of the art for cellular networks. The second part contains a detailed overview of URLLC use cases and the requirements that must be fulfilled by the cellular networks to support them. The work in this thesis is done as part of a collaboration project between IRIMAS lab in Université de Haute-Alsace, France and Institute for Reliable Embedded Systems and Communication Electronics (ivESK) in Offenburg University of Applied Sciences, Germany. The selected use cases of URLLC are part of the research interests of both partner institutes. The third part presents a detailed study and evaluation of user- and control-plane latency mechanisms in current generation of cellular networks. The evaluation and analysis of these latencies, performed with the open-source ns-3 simulator, were conducted by exploring a broad range of parameters that include among others, traffic models, channel access parameters, realistic propagation models, and a broad set of cellular network protocol stack parameters. These simulations were performed with low-power, low-cost, and wide-range devices, commonly called IoT devices, and standardized for cellular networks. These devices use either LTE-M or Narrowband-IoT (NB-IoT) technologies that are designed for connected things. They differ mainly by the provided bandwidth and other additional characteristics such as coding scheme, device complexity, and so on.
The fourth part of this thesis shows a study, an implementation, and an evaluation of latency reduction techniques that target the different layers of the currently used Long Term Evolution (LTE) network protocol stack. These techniques based on Transmission Time Interval (TTI) reduction and Semi-Persistent Scheduling (SPS) methods are implemented into the ns-3 simulator and are evaluated through realistic simulations performed for a variety of low-latency use cases focused on industry automation and vehicular networking. For testing the proposed latency reduction techniques in cellular networks, since ns-3 does not support NB-IoT in its current release, an NB-IoT extension for LTE module was developed. This makes it possible to explore deployment limitations and issues.
In the last part of this thesis, a flexible deployment framework called Hybrid Scheduling and Flexible TTI for the proposed latency reduction techniques is presented, implemented and evaluated through realistic simulations. With help of the simulation evaluation, it is shown that the improved LTE network proposed and implemented in the simulator can support low-latency applications with low cost, higher range, and narrow bandwidth devices. The work in this thesis points out the potential improvement techniques, their deployment issues and paves the way towards the support for URLLC applications with upcoming cellular networks.
Due to the rapidly increasing storage consumption worldwide, as well as the expectation of continuous availability of information, the complexity of administration in today’s data centers is growing permanently. Integrated techniques for monitoring hard disks can increase the reliability of storage systems. However, these techniques often lack intelligent data analysis to perform predictive maintenance. To solve this problem, machine learning algorithms can be used to detect potential failures in advance and prevent them. In this paper, an unsupervised model for predicting hard disk failures based on Isolation Forest is proposed. Consequently, a method is presented that can deal with the highly imbalanced datasets, as the experiment on the Backblaze benchmark dataset demonstrates.
In the work at hand, we state that privacy and malleability of data are two aspects highly desired but not easy to associate. On the one hand, we are trying to shape data to make them usable and editable in an intelligible way, namely without losing their initial information. On the other hand, we are looking for effective privacy on data such that no external or non-authorized party could learn about their content. In such a way, we get overlapping requirements by pursuing different goals; it is trivial to be malleable without being secure, and vice versa. We propose four “real-world” use cases identified as scenarios where these two contradictory features are required and taking place in distinct environments. These considered backgrounds consist of firstly, cloud security auditing, then privacy of mobile network users and industry 4.0 and finally, privacy of COVID-19 tracing app users. After presenting useful background material, we propose to employ multiple approaches to design solutions to solve the use cases. We combine homomorphic encryption with searchable encryption and private information retrieval protocol to build an effective construction for the could auditing use case. As a second step, we develop an algorithm to generate the appropriate parameters to use the somewhat homomorphic encryption scheme by considering correctness, performance and security of the respective application. Finally, we propose an alternative use of Bloom filter data structure by adding an HMAC function to allow an outsourced third party to perform set relations in a private manner. By analyzing the overlapping bits occurring on Bloom filters while testing the inclusiveness or disjointness of the sets, we show how these functions maintain privacy and allow operations directly computed on the data structure. Then, we show how these constructions could be applied to the four selected use cases. Our obtained solutions have been implemented and we provide promising results that validate their efficiency and thus relevancy.
Among the billions of smartphone users in the world, Android still holds more than 80% of the market share. The applications which the users install have a specific set of features that need access to some device functionalities and sensors that may hold sensitive information about the user. Therefore, Android releases have set permission standards to let the user know what information is being disclosed to the application. Along with other security and privacy improvements, significant changes to the permission scheme are introduced with the Android 6.0 version (API level 23). In this master thesis, the Android permission scheme is tested on two devices from different eras. The evolution of Android over the years is examined in terms of confidentiality. For each device, two applications are built; one focused on extracting every piece of information within the confidentiality scope with every permission declared and/or requested, and the other app focused on getting this type of information without user notification. The resulting analysis illustrates whether how and in what way the Android permission scheme declined or improved over time.
In this work, an implementation of the somewhat homomorphic BV encryption scheme is presented. During the implementation, care was taken to ensure that the resulting program will be as efficient as possible i.e. fast and resource-saving. The basis for this is the work of Arndt Bieberstein, who implemented the BV scheme with respect to functionality. The presented implementation supports the basics of the BV scheme, namely (symmetric and asymmetric) encryption, decryption and evaluation of addition as well as multiplication. Additionally, it supports the encoding of positive and negative numbers, various gaussian sampling methods, basically infinitely large polynomial coefficients, the generation of suitable parameters for a use case, threading and relinearization to reduce the size of a ciphertext after multiplications. After presenting the techniques used in the implementation, it’s actual efficiency is determined by measuring the timings of the operations for various parameters.
The identification of vulnerabilities is an important element of the software development process to ensure the security of software. Vulnerability identification based on the source code is a well studied field. To find vulnerabilities on the basis of a binary executable without the corresponding source code is more challenging. Recent research has shown how such detection can be performed statically and thus runtime efficiently by using deep learning methods for certain types of vulnerabilities.
This thesis aims to examine to what extent this identification can be applied sufficiently for a variety of vulnerabilities. Therefore, a supervised deep learning approach using recurrent neural networks for the application of vulnerability detection based on binary executables is used. For this purpose, a dataset with 50,651 samples of 23 different vulnerabilities in the form of a standardised LLVM Intermediate Representation was prepared. The vectorised features of a Word2Vec model were then used to train different variations of three basic architectures of recurrent neural networks (GRU, LSTM, SRNN). For this purpose, a binary classification was trained for the presence of an arbitrary vulnerability, and a multi-class model was trained for the identification of the exact vulnerability, which achieved an out-of-sample accuracy of 88% and 77%, respectively. Differences in the detection of different vulnerabilities were also observed, with non-vulnerable samples being detected with a particularly high precision of over 98%. Thus, the methodology presented allows an accurate detection of vulnerabilities, as well as a strong limitation of the analysis scope for further analysis steps.