Abstract 
Runtime verification offers scalable solutions to improve the safety and reliability of systems. However, systems that require verification or monitoring by a third party to ensure compliance with a specification might contain sensitive information, causing privacy concerns when usual runtime verification approaches are used. Privacy is compromised if protected information about the system, or sensitive data that is processed by the system, is revealed. In addition, revealing the specification being monitored may undermine the essence of third-party verification. 

In this work, we propose a novel protocol for the privacy-preserving runtime verification of systems against formal sequential specifications, which ensures that the system remains oblivious to the monitored specification, while the monitor learns only whether the system satisfies the specification and nothing more. Our protocol adapts and improves existing techniques used in cryptography, and more specifically, two-party computation.

Bio 
Mahyar is a PhD student at IST Austria, supervised by Tom Henzinger. His research focus is on the intersection of runtime verification and private computing; specifically, his work is on designing private schemes for runtime verification, using techniques from cryptography. Before starting his graduate studies, Mahyar got his BSc in computer engineering from the University of Tehran, in 2023. 

More: https://mahykari.github.io
Photo provided by speaker

Abstract
Machine learning is becoming ubiquitous, influencing critical decisions across multiple sectors. However, most ML systems are one-model-fits-all solutions that are trained once with occasional updates, limiting their adaptability and performance over time. Continual ML approaches provide an adaptable solution that can handle distribution shifts, cater to rare events, and maintain robustness against adversarial attacks.

This talk explores the methodologies and applications of continual learning, emphasizing its importance in various domains such as automotive, robotics, and natural language processing (NLP). In the automotive domain, we explore incremental learning approaches for adapting gesture recognition systems to individual drivers, enhancing human-vehicle interaction. In robotics, we combine reinforcement and imitation learning to develop a curriculum learning approach to create a surgeon-in-the-loop ophthalmic robotic apprentice. In the NLP domain, we teach LLM agents to learn from their mistakes, adapt to contextually valid changes, and protect against privacy, security, and social engineering attacks.

Bio 
Amr Gomaa is a final-year Ph.D. candidate in Human-centered AI & Applied ML at the German Research Center for Artificial Intelligence (DFKI). He is also a Visiting Researcher at the University of Cambridge, focusing on feasible ML (specifically LLMs) solutions for HCI design processes and interfaces, and ensuring the robustness of contextual LLM agents in collaboration with Microsoft Cambridge. His work focuses on developing adaptable machine learning systems, including incremental learning, reinforcement learning, and imitation learning, applied to domains such as automotive, robotics, health, and LLM agents.

His work is published in top-tier conferences such as NeurIPS, IROS, and IUI. Amr has led multiple research projects in collaboration with industry partners such as BMW and Zeiss, and secured multiple research grants for several projects such as “Secure Language Models for Knowledge Management” and “Hybrid Reinforcement Learning and Imitation Learnin

More: amrgomaaelhady.github.io

Photo provided by speaker

Abstract

PhD student Lena Heimberger visited Cloudflare in Lisbon from July to October. Now that she is back, she will talk about Transparency, the Certificate ecosystem, Merkle Tree Certificates, fallback strategies, protocols and why Portugese coffee culture is better than Italian. More concretely, she studied the certificate ecosystem in the context of post-quantum certificates, trying to sketch the design space for fallback mechanisms for so-called slow-issuance certificates. The result is an upper bound for fallback probabilities that can be used to study methods for fallbacks. As a part of this study, Lena and her colleagues found that SCT auditing is on the brink of being practical when using Private Information Retrieval methods.

Parts of the talk have already been presented at transparency.dev in London last month, and will be submitted in form of a paper soon.

The talk is aimed at Master’s students, but anybody is welcome. Questions, also about the internship, are explicitly encouraged.

 

At the event we present our new open bachelor’s thesis (and master’s thesis) topics and award prizes to excellent students.  

If you’re interested in joining us for your bachelor’s thesis in security, this is the best way to get an impression of our topics as well as how a bachelor’s thesis at IAIK works: You’ll hear about our research areas and current hot topics, our Bachelor@IAIK program where you can work on your thesis together with your fellow students in one of our offices if you like, and maybe you’ll get to know your supervisor while chatting along.

THE AWARDS:

IAIK Student Research Excellence Award: Students at IAIK who became co-authors of a scientific publication in the context of a thesis or project receive this award.

IAIK Bachelor Excellence Award: This award is for students majoring in “Information Security” at TU Graz and who have completed their bachelor’s degree with distinction at TU Graz. Application deadline: Oct 9th 2024!

The event will also be the kick-off lecture in Introduction to Scientific Working (ISW) where you will be able to choose your preferred topic!   

We are looking forward to meeting you!

 

Abstract
Sequential decision-making tasks often require satisfaction of multiple, partially-contradictory objectives. Existing approaches are monolithic, where a single policy fulfills all objectives. I will present auction-based scheduling, a new modular framework for multi-objective sequential decision-making. In this framework, each objective is fulfilled using a separate and independent policy. The policies are then composed through a novel run-time composition mechanism, where at each step, they need to bid from pre-allocated budgets for the privilege of choosing the next action. The bidding encourages the policies to adjust their bids according to their urgencies to act, and whoever bids the highest gets scheduled first. We study the following decentralized synthesis problem: How to compute bidding-equipped policies whose composition will simultaneously fulfill all objectives? I will present solutions of the decentralized synthesis problem for path planning on finite graphs with two temporal objectives.

Bio
Kaushik Mallik is a postdoctoral researcher at the Institute of Science and Technology Austria (ISTA). His research interests are broadly in formal verification and synthesis of reactive software. He received the 2023 ETAPS Doctoral Dissertation Award, nominations for best paper awards in HSCC and TACAS, and a number of merit-based fellowships during his bachelor’s and master’s studies. He holds B. Tech (2012) in Electrical Engineering from Meghnad Saha Institute of Technology (India), M. Tech (2015) in System and Control from IIT Roorkee (India), and Dr. rer. nat. (PhD, 2022) in Computer Science from RPTU Kaiserslautern (Germany).

Photo provided by speaker