News
Next On-Site Seminar on 27.05.2026, CISPA C0 Room 0.02
Written on 21.05.2026 09:56 by Xinyi Xu
Dear All,
The next seminar(s) will take place on 27.05.2026, 14:00 - 16:00, CISPA C0 (Stuhlsatzenhaus 5, 66123 Saarbrücken) - C0 Room 0.02. Presenters and their advisors are encouraged to present in person. We especially encourage other students and teachers to attend and present in person as well.
For presenters,
1. We would book the room half an hour in advance, so you are encouraged to arrive a few minutes early to set up your own poster.
2. For this session, you need to print the poster on your own. The size of the poster should be 116x86cm or 86x116cm. You can use the poster printing service of Saarland University (https://www.uni-saarland.de/en/page/uds-card/functions/printing.html -> Posterdruck A0).
3. You need to present your poster in a much smaller group, but you are encouraged to roam around and ask questions about other posters.
4. We encourage you to bring your laptop to present your demo; there will be small tables in the room where you can put your laptop.
Presenters: Marija Jovanovikj, Christina Subedi, Sai Suresh Macharla Vasu
27.05.2026, 14:00 - 16:00, CISPA C0 (Stuhlsatzenhaus 5, 66123 Saarbrücken)
Presenter: Marija Jovanovikj
Type of Poster: Master Intro
Advisor: Dañiel Gerhardt, Katharina Krombholz
Title: User-Centered Security and Privacy Enhancements in Smart Home Dashboard Design
Research Area: RA6: Empirical and Behavioural Security
Abstract: Smart home users typically own several smart home devices, therefore providing proper device management is crucial. A central control hub dashboard provides an interface for monitoring and manipulating of smart devices from a single location, as well as visualization of the overall state of the smart home environment. Yet, existing dashboards lack security and privacy features. Security-wise, many smart home devices come preconfigured with default settings, which do not provide strong enough protection. Furthermore, the compromise of one poorly configured device can result in lateral movement, compromising further devices on the same network. Privacy-wise, smart home devices measure and collect data, which is eventually sent to remote servers for further diagnosis and analytics. Although the collected data seems harmless, it can still leak patterns of a user’s lifestyle, such as when the user is at home, what is the user doing and the interval duration of some action. A security and privacy enhanced dashboard can be beneficial for both IT expert and non-IT expert users. For IT expert users, the centralized configuration will help them perform configurations quickly and save time. For non-IT expert users, the benefits will be greater, as it will help raise their awareness about the possible real-time vulnerabilities of their overall smart home environment through visual warnings, as well as provide tips how to strengthen the overall security. Furthermore, the privacy features will promote personal customization, giving users the option to limit and restrict the information that will be shared according to their preferences. Our aim is to develop a dashboard prototype that includes the security and privacy functionalities, according to previously conducted research guidelines regarding users' preferences. The completed prototype will be tested by both IT and non-IT expert users. The collected and analyzed information will be helpful for confirming the previously conducted related work and set foot for future work, bringing this security and privacy enhanced smart home dashboard concept closer to actual practical implementation.
27.05.2026, 14:00 - 16:00, CISPA C0 (Stuhlsatzenhaus 5, 66123 Saarbrücken)
Presenter: Christina Subedi
Type of Poster: Master Intro
Advisor: Jonas Hielscher
Title: LLM-Based Analysis of US SEC 10-K Filings
Research Area: RA6: Empirical and Behavioural Security
Abstract: Form 10-K is an annual report mandated by the U.S. Securities and Exchange Commission (SEC) that provides a comprehensive overview of a company's financial performance and business operations. Beginning in December 2023, the SEC introduced Item 1C: Cybersecurity, requiring publicly listed companies to disclose material cybersecurity incidents, risk management strategies, and the role of executive management and the board of directors in overseeing cybersecurity. This thesis systematically investigates the cybersecurity strategies and governance structures adopted by U.S. companies in response to this new disclosure requirement. Drawing on 10-K filings from 2024 and 2025, we conduct a large-scale analysis examining cybersecurity reporting hierarchies, organizational governance frameworks, and year-over-year structural changes between the two periods. The analysis is carried out using PageIndex, a large language model (LLM)-based tool designed to extract and interpret structured information from regulatory filings at scale.
27.05.2026, 14:00 - 16:00, CISPA C0 (Stuhlsatzenhaus 5, 66123 Saarbrücken)
Presenter: Sai Suresh Macharla Vasu
Type of Poster: Master Intro
Advisor: Ivaxi Sheth, Philipp Christmann, Ruta Binkyte, Mario Fritz
Title: Long-Horizon Personalization via Personal Knowledge Graphs
Research Area: RA2: Trustworthy Information Processing
Abstract: Large Language Models are increasingly moving from stateless systems toward memory-enabled assistants that retain user information across interactions. However, current memory systems often lack temporal awareness, structured organization, and mechanisms to handle evolving user preferences over long periods of time. In this thesis, we investigate long-horizon personalization through Personal Knowledge Graphs (PKGs), where user memories are represented as structured, temporal, and stability-aware facts. We propose an agentic memory architecture that supports continual memory updates and enables controlled retrieval of relevant user information during interaction. The work explores how structured and agent-driven memory systems can improve personalization quality while reducing outdated, redundant, or noisy memories in long-term conversational systems. In particular, the system employs an agentic pipeline that dynamically decides whether newly observed information should be added, updated, deprecated, or ignored based on conversational context and existing memory state.