Themen

Artificial Intelligence

✘ StarCraft AI Research – English only

Real-Time Multiplayer Games are regarded as one of the final frontiers for game AIs. While chess and recently also Go can nowadays be reliably won by machines, in computer games - such as MOBAs, or the strategy games StarCraft and StarCraft II - the top human players are still considered better at the game than AIs. However, researches are working on changing this. In 2017, DeepMind partnered with the game producer Blizzard, to provide a StarCraft II API that allows the testing of AIs. This seminar should review and present the most relevant AI research around the games of StarCraft and StarCraft II.

✘ Generative Adversarial Networks (GANs) – English only

AI nowadays is getting better at identifying things with high accuracy allowing, e.g., the dream of autonomous driving to become a reality. However, this high accuracy needs, generally speaking, a huge set of training data. Instead of having a real-world dataset, which might be costly to have in many cases, a recent approach proposed to use a generated synthetic data instead. This is done through the so-called “Generative Adversarial Networks (GANs)”. The goal is then to create ultra-realistic e.g. images or sounds to train the network on. This, in fact, requires giving the machine a sense of imagination. In this seminar, you will investigate the latest trends in the field of GANs while reporting recent results found in literature. Moreover, you will address what challenges face GANs including security ones. 

Distributed Systems

✘ Resource Scheduling in Cloud Computing – English only

With increased popularity of Cloud Computing the approach of treating multiple nodes as one big resource unit came up. This allows to run multiple different applications on one cluster at the same time. The biggest challenge is to schedule the processes of the applications without overstress the cluster or slowing down one of the applications.

The goal of this seminar is to look at the different scheduling approaches and their use caces in cloud computing.

Eugen Frasch

✘ Serverless Computing – English only

Serverless computing it the next level of abstraction in cloud computing, with the goal that customers do not have to worry about operational concerns and can completely focus on their code. In contrast to the pre-purchased payment model of Infrastructure as a Service (IaaS) and Plattform as a Service (PaaS), serverless computing features a pay-per-use model, where a customer is only charged for the resources that were actually used. Despite what the name suggests, there are still servers involved which are managed by the cloud provider, allowing the provider to dynamically allocate resources based on the customer's demands. The primary task of this seminar is an introduction of the overall topic and a comprehensive survey of existing open- and closed-source solutions of serverless computing.

✘ Modern Peer-to-Peer Topology Trends – English only

Peer-to-peer networks are used for many applications, e.g., sharing of larger data items, data dissemination in crypto currencies and many more. Many advances have been made for structured and unstructured networks on how to build and maintain the network. The goal of this seminar is to look at several modern networks and examine their topology, routing principles and methods to maintain the topology.

David Mödinger

Cryptography

✘ Range Proofs – English only

Confidential transactions in crypto currencies require range proofs to detect integer overflows. Any output amount of a transaction has to be a positive integer. As storage is valuable on block-chains, the goal is to reduce the size as much as possible. Recent advances in bulletproofs reduce the size significantly. The paper should compare the different existing methods and point out how the improvements are achieved.

✘ Ring Signatures – English only

Common signatures require one private key and the corresponding public key and provide authenticity for the signed data. Ring signatures allow to disguise the real signer in a set of public keys. The verifier can only check that the signer was in possession of at least one corresponding private key, but not which one. In comparison to group signatures, the key advantage is that ring signatures can be created without the interaction of the decoy participants. The paper should give an overview of the different constructions of ring signature schemes and an in-depth explanation of one of them.

Security Algorithms

✘ Secure Multiparty Computation – English only

Secure multiparty computation (MPC) enables several parties, each with a private input, to compute a joint function without exposing their input to each other. This attracts many application scenarios, e.g. training a machine learning model on private input from several companies/hospitals, where privacy is crucial. Within the last decade, the development of MPC has been on the practical side and many efficient implementations have been proposed. In this seminar, you will investigate those practical implementations of secure MPC while addressing the features of each including the suggested functionality, and check whether there is any real-world usage of each. Then, selected implemetations are to have in-depth comparison among each other.

✘ Risk Evaluation Methodologies – English only

The Common Vulnerability Scoring System is a well-known methodology for determining how critical a vulnerability is. Your task is to find and compare other methodologies/ systems/ methods for assessing how much risk a security bug will have.

✘ IDS Algorithms – English only

The Aho–Corasick is a well-known algorithm used in Intrusion Detection Systems for finding strings in a stream of characters. This can be used for example in a firewall to search for certain words in the packages and if a match is found, raise an alert. Your task is to look at existing IDS solutions, find out what algorithms they use and compare them to Aho-Corasick in terms of speed and memory consumption. You don’t need to go into details about each algorithm and understand every aspect of it, but enough how they will perform. If you find to many algorithms, just take 5 and compare them.

Michael Wolf

Network Security

✘ The Dark Side of Network Data Analysis: Censorship Infrastructure Around the World – English only

From the Big Firewall of China to The Supreme Council of Virtual Space in Iran, authoritarian regimes around the world try to limit personal freedom to suppress possible opposition or – so they say – protect their citizens from outside influence or "moral decline". The goal of this seminar is to analyse censorship infrastructure around the world on a technical level and to discuss the societal impact thereof.

✘ Analysis of Modern Network Testing Approaches – English only

Every (new) network protocol (e.g., TCP or NetFlow) and device (e.g., switches or routers) needs to be tested. The main task of this seminar is to outline different modern testing approaches. In what way do researches test network protocols and devices. In what network environment is the protocol/device tested? What traffic is used? How often are test runs repeated? The seminar paper should outline modern approaches and state the pros and cons of the presented methodologies.

✘ Network Security Breaches – English only

The goal of this seminar is the outlining of popular network security breaches (2-3 examples).  Subsequently, state-of-the-art protection or detection approaches against these presented breaches should be explained.

✘ What did we learn in almost 20 years of TLS? – English only

In 2018, the latest version of TLS - TLS 1.3 - was finally finished and published. The discussions and decisions that were made during the development phase are interesting to follow. In an attempt to design the most secure version of TLS that is nowadays possible, the IETF working group decided to deprecate and remove quite some parts of older TLS releases that have a rather troubling history. As part of this work, we want to follow the development of TLS 1.3 and try to learn from it for the design of cryptographic (network) protocols in general. Towards this goal, we follow two questions. What do we learn about protocol design by understanding the removed parts in TLS 1.3? It defines a new key agreement protocol, removes old cryptographic algorithms and changes the defaults. What are the underlying problems that lead to these decisions? What can we learn when designing cryptographic protocols on our own? What can we learn from the new features in TLS 1.3? What is the general direction in which the community is moving? Which security aspects have grown more important nowadays? Which building blocks are considered generally better than others?

Clemens Lang (BMW Car IT)

Internet of Things

✘ Individual Mobile Sensing in Healthcare - Trends and State-of-the-art – English only

Individual Mobile Sensing or Personal Mobile Sensing often focuses on the aspects of sensor data collection and analysis applied specifically for the pruposes of education, diagnosis, event detection, treatment, or monitoring. The aim of this seminar is to study recent trends and review state-of-the-art already reported in existing literature, specifically applied in the domain of healthcare.

Muntazir Mehdi

✘ Electroencephalography (EEG) using Smartphones – English only

Electroencephalography (EEG) is the method of monitoring the electrical activity of the brain. Mobile EEGs are dedicated hardware equipment capable of coupling with many commonplace smartphones. In this seminar report, the students are required to survey the current state of EEGs and mobile EEG systems that can be coupled with smartphones. The students would be further required to study various applications of EEG systems.

Muntazir Mehdi

✘ Smart Home Security – English only

Home automation has been a common theme of science fiction for a long time but has recently become reality. Newly constructed homes often come with pre-installed smart home features based on proprietary hardware/software and the same can be retrofitted to older buildings. Additionally, a huge community of DIY enthusiasts develops open source and open hardware alternatives. However, security is often not the main priority of these products or can be difficult to implement correctly, due to hardware and power constraints. Numerous examples, such as smart door and garage locks, show that this is becoming a serious issue. This seminar should outline the differences in conventional IT security and IoT security and survey the security of common home automation software (e.g., home-assistant), protocols (e.g., MQTT), and devices (e.g., ESP8266).

✘ IoT Security - State of the Union  – English only

The Internet of Things is an emerging field of Internet-enabled embedded devices. Applications range from connected factory lines to large scale sensor networks to medical equipment. Most of the applications share the limitations of classical embedded systems. They run on battery and have limited computational resources. Yet, they often control physical systems such as industry robots, children's toys or even critical devices such as infusion pumps in hospitals. Internet of Things devices have a bad reputation in the security research community. In this work, we want to get an overview of the specialties and problems of IoT devices. You should introduce the general setup and explain problems based on a scientific case study of your choice. What are the answers of the security community? How is research trying to improve the general situation in this area? Which are the most important issues that should be addressed?

Martin Lang (BMW Car IT)

Programming Languages

✘ The Julia Programming Language – English only

Since its inception almost 10 years ago, Julia has grown to be a full-fledged programming language for scientific computing, with an impressive list of features. It has in recent months been touted as a successor to R and an alternative to MatLab, and aims to provide the scientific community with an easy-to-learn language while coming close to the speed of C.

This seminar should provide an introduction to the Julia language, describe its main features, and include examples of how it can be used both in the context of general and scientific computing.

Gerhard Habiger