News
Next Seminar on 13.03.2024
Written on 11.03.2024 16:04 by Mang Zhao
Dear All,
The next seminar(s) take place on 13.03.2024 at 14:00 (Session A) and 14:00 (Session B).
Session A: (14:00-15:30)
Lenny Händler, Lennard, Niklas Britz
https://cispa-de.zoom.us/j/96786205841?pwd=M3FOQ3dSczRabDNLb3F1czVXVUpvdz09
Meeting-ID: 967 8620 5841
Kenncode: BT!u5=
Session B: (14:00-15:30)
Dominik Sautter, Jannis, Houcai Li
https://cispa-de.zoom-x.de/j/66136901453?pwd=YVBSZU9peUpvUlk4bWp3MDR4cGlUUT09
Session A:
14:00 - 14:30
Speaker: Lenny Händler
No Information is provided.
14:30 - 15:00
Speaker: Lennard Tworeck
Type of talk: Bachelor Outro
Advisor: Robert Künnemann, Kevin Morio
Title: A parser for the spthy protocol modeling language
Research Area: RA2
Abstract:
Security protocols ensure the confidentiality, integrity and authentication of transmitted data between two or more parties and are the basis of secure communication. Proving the accuracy of these protocols is an error-prone and arduous task, which is why protocol verifiers have been developed to automate this task. Tamarin is such a protocol verifier and allows for symbolic modeling and analysis of security protocols. To use Tamarin, the user defines the protocol, environment and security properties in a security protocol model, which is specified in the so-called security protocol theory (spthy) file format developed especially for Tamarin.
In this thesis, an independent parser was created that allows parsing such files for third party tools or Tamarin extensions. A parser is a software component that converts
textual input into a machine-readable format based on a formal grammar. The spthy parser was developed with the objective of flexibility, minimal dependencies, and support for various programming languages such as Python, Go, Rust and Haskell. The tree-sitter parser generator was employed to create the parser. The process involved
to first define a grammar, accepting the syntax of spthy files and utilizing this grammar as an input to the parser generator. While the syntax of the file format was already
partially documented in the Tamarin manual, a complete description was missing. The parser generator approach did not just build a parser for spthy files, but also documented
its syntax.
The primary aim was to create a versatile tool that facilitates the development of Tamarin extensions. The result is a reusable parser that not only enhances the ease of creating extensions for Tamarin, but also offers comprehensive syntax documentation for the spthy file format.
15:00 - 15:30
Speaker: Niklas Britz
Type of talk: Bachelor Intro
Advisor: Dr. Nico Döttling
Title: Gaussian Linear Secret Sharing
Research Area: 2
Abstract:
Secret sharing is a cryptographic technique to distribute a secret among different parties. Only a specified amount of the parties can reconstruct the secret together, while smaller party sizes learn nothing or little about the original secret when combining their respective information.
In this work, we present two threshold secret sharing schemes. The first is a continuous linear secret sharing scheme over the group R. We prove the correctness and security of this scheme.
The second is an approximate linear scheme over Z, for which we prove correctness. Here, approximate linear means that the scheme is linear up to some minor error.
Using Gaussian distributions for secret sharing over R and Z, we present a new approach to linear secret sharing.
Session B:
14:00 - 14:30
Speaker: Dominik Sautter
No Information is provided.
14:30 - 15:00
Speaker: Jannis Cavelius
Type of talk: Bachelor Intro
Advisor: Prof. Sebastian Stich
Title: Hyperparameter Optimization in Federated Learning using Covariance matrix adaptation
RA2: Reliable Security Guarantees
Abstract:
Federated Learning(FL) has risen in popularity over the last few years. It can be split into two main components. First, a central server, with the global model, that controls the training, and second, multiple clients (i.e. mobile devices or companies), that perform local updates to the global model using their respective data. At the end, only the updates to the model weights are shared with the server and used to update the global model.
The decentralized training of the model, which results in high data privacy, can lead to bottlenecks either in the form of communication or computation, i.e., a bad internet connection or old hardware, and slow down the training.
Hyperparameter Optimization, which needs lots of evaluations to find a good configuration, is difficult, as these might not be possible either through privacy restrictions or the bottlenecks mentioned above.
Federated Learning models therefore often use the hyperparameters, which are predefined by the optimizers, to save time and resources.
This thesis tries to use the evolutionary strategies Covariance Matrix Adaptation to efficiently find the models Hyperparameters. The approach is then compared to baselines, such as Grid- and Random search, on various datasets and problem types.
Another goal of my work is to bring back gradient free optimizers into the spotlight and make space for new work and research.
15:00 - 15:30
Speaker: Houcai Li
Type of talk: Master Intro
Advisor: Dr. Dominic Steinhoefel
Title: XML-Schema Refinement: A Human-in-the-Loop Approach
Research Area: RA3
Abstract:
XML, the Extensible Markup Language, provides detailed information on both content and the structure of data. The structure and the set of vocabularies of XML documents can be declared in an XML schema, which can be used to validate XML documents.
Therefore, XML schemas can play an important role in the field of software security as it filters out unwanted input when properly defined. It also has the potential to be used as specifications for automated software testing campaigns.
Since many XML schemas are only laxly defined and can therefore present security risks, we present a GUI to refine a schema based on available XML documents conforming to it. Built on EditiX, a public open-source XML editor, the presented GUI is able to analyze XML documents using the (W3C) XML schema currently displayed, and mark the positions where suggestions are available for refining the schema.