Friday 26.08.2022 Detailed

Explainable Empirical Risk Minimization

Alexander Jung (Assistant Professor, Aalto University, Finland; Associate Editor, IEEE Signal Processing Letters)

Abstract: The successful application of machine learning (ML) methods becomes increasingly dependent on their interpretability or explainability. Designing explainable ML systems is instrumental to ensuring transparency of automated decision-making that targets humans. The explainability of ML methods is also an essential ingredient for trustworthy artificial intelligence. A key challenge in ensuring explainability is its dependence on the specific human user (“explainee”).
The users of machine learning methods might have vastly different background knowledge about machine learning principles. One user might have a university degree in machine learning or related fields, while another user might have never received formal training in high-school mathematics. We measure explainability via the conditional entropy of predictions, given some user signal. This user signal might be obtained from user surveys or biophysical measurements.
We propose explainable empirical risk minimization (EERM) principle of learning a hypothesis that optimally balances between the subjective explainability and risk.
The EERM principle is flexible and can be combined with arbitrary machine learning models. We present several practical implementations of EERM for linear models and decision trees. Numerical experiments demonstrate the application of EERM to detecting the use of inappropriate language on social media.

Improved Integer-wise Homomorphic Comparison and Division based on Polynomial Evaluation

Koki Morimura, Daisuke Maeda and Takashi Nishide (University of Tsukuba, Japan)

Machine-Learning Side-Channel Attacks on the GALACTICS Constant-Time Implementation of BLISS

Soundes Marzougui (Technical University of Berlin, Germany), Nils Wisiol (Technical University of Berlin, Germany), Patrick Gersch (Technical University of Berlin, Germany), Juliane Krämer (University of Regensburg, Germany) and Jean-Pierre Seifert (Technical University of Berlin, Germany)

ZEKRO: Zero-Knowledge Proof of Integrity Conformance

Heini Bergsson Debes (Technical University of Denmark, Denmark) and Thanassis Giannetsos (Ubitech Ltd., Greece)

Comparing machine learning correlations to domain experts’ causal knowledge: Employee turnover use case

Eya Meddeb (Department of Computing, Worcester Business School, University of Worcester, United Kingdom), Christopher Bowers (Department of Computing, Worcester Business School, University of Worcester, United Kingdom), Lynn Nichol (Worcester Business School, University of Worcester, United Kingdom)

Machine learning and knowledge extraction to support work safety for smart forest operations

Ferdinand Hönigsberger (University of Natural Resources and Life Sciences, Vienna, Austria), Anna Saranti (University of Natural Resources and Life Sciences, Vienna, Austria), Alessa Angerschmid (University of Natural Resources and Life Sciences, Vienna, Austria), Carl Orge Retzlaff (University of Natural Resources and Life Sciences, Vienna, Austria), Christoph Gollob (University of Natural Resources and Life Sciences, Vienna, Austria), Sarah Witzmann (University of Natural Resources and Life Sciences, Vienna, Austria), Arne Nothdurft (University of Natural Resources and Life Sciences, Vienna, Austria), Peter Kieseberg (University of Applied Sciences St.Poelten, Austria), Andreas Holzinger (University of Natural Resources and Life Sciences, Vienna, Austria) and Karl Stampfer (University of Natural Resources and Life Sciences, Vienna, Austria)

Security for Healthcare Services: Needs, Solutions and Challenges

Prof. George Spanoudakis, City University London

Abstract: This talk reviews the current state of practice and state of the art in the security of healthcare services. More specifically, it reviews the key security challenges faced by healthcare service providers, the types of security assessments needed, the methods for security risk management, and the landscape of the security solutions available. The latter are reviewed in terms of maturity and the expectations for emerging solutions in the short (1-2 years) and medium-term (3-5 years).

Register Automata for Malware Specification

Tayssir Touili (LIPN, CNRS & University Paris 13, France)

CopypastaVulGuard – A browser extension to prevent copy and paste spreading of vulnerable source code in forum posts

Holger Schmidt, Max van Aerssen, Christian Leich, Abdulkader Benni, Salar Al Ali and Jakob Tanz (Düsseldorf University of Applied Sciences, Germany)

A Quantitative Assessment of the Detection Performance of Web Vulnerability Scanners
Emma Lavens, Pieter Philippaerts, Wouter Joosen (imec – DistriNet, KU Leuven, Belgium)

Symbolic analysis meets federated learning to enhance malware identifier
Charles-Henry Bertrand Van Ouytsel, Khanh Huu The Dam and Axel Legay (Universite Catholique de Louvain, Belgium)

SoK: A Systematic Literature Review of Knowledge-Based Authentication on Augmented Reality Head-Mounted Displays

Reyhan Düzgün (Karlsruhe Institute of Technology, Germany), Naheem Noah (University of Denver, United States), Peter Mayer (Karlsruhe Institute of Technology, Germany), Sanchari Das (University of Denver, United States) and Melanie Volkamer (Karlsruhe Institute of Technology, Germany)

Combining Variational Autoencoders and Transformer Language Models for Improved Password Generation

David Biesner (Fraunhofer IAIS and University of Bonn, Germany), Kostadin Cvejoski (Fraunhofer IAIS, Germany) and Rafet Sifa (Fraunhofer IAIS, Germany)

Analysis and Evaluation of Hardware Trust Anchors in the Automotive Domain

Christian Plappert (Fraunhofer Institute SIT, ATHENE, Germany), Andreas Fuchs (Fraunhofer Institute SIT, ATHENE, Germany) and Ronald Heddergott (CARIAD SE, Germany)

Multi-Account Dashboard for Authentication Dependency Analysis

Daniela Pöhn (Universität der Bundeswehr München, FI CODE, Germany), Nils Gruschka (University of Oslo, Norway) and Leonhard Ziegler (Universität der Bundeswehr München, Germany)

Multi-label Classification of Cybersecurity Text with Distant Supervision

Masahiro Ishii, Kento Mori, Ryoichi Kuwana and Satoshi Matsuura (Tokyo Institute of Technology, Japan)

Risk Assessments Considering Safety, Security, and Their Interdependencies in OT Environments

Siegfried Hollerer (TU Wien, Austria), Thilo Sauter (TU Wien, Danube Univ., Austria), Wolfgang Kastner (TU Wien, Austria)

Improving Network, Data and Application Security for SMEs

Christos Tselios (University of Patras, Greece), Ilias Politis (University of Piraeus, Greece), Christos Xenakis (University of Piraeus, Greece)

Android Permission Manager, Visual Cues, and their Effect on Privacy Awareness and Privacy Literacy

Vera Schmitt (Technische Universität Berlin, Germany), Maija Poikela (Fraunhofer Institute for Applied and Integrated Security, Germany), Sebastian Möller (Technische Universität Berlin, Germany)

Design Space Exploration of DICE

Dominik Lorych and Lukas Jäger (Fraunhofer Institute SIT | ATHENE, Germany)

DevSecOps In Embedded Systems An Empirical Study Of Past Literature

Hasan Yasar and Sam E Teplov (Carnegie Mellon University, United States)

MetaSEnD A Security Enabled Development Life Cycle Meta-Model

Daniele Granata, Massimiliano Rak and Giovanni Salzillo (Università della Campania Luigi Vanvitelli, Italy)

Safety and Security Analysis using LDA based on Case Reports Case Study and Trust Evaluation Method

Katsuyuki Umezawa (Shonan Institute of Technology, National Institute of AIST), Hiroki Koyanagi (Internet Initiative Japan Inc.), Sven Wohlgemuth (SECOM Co., Ltd.), Yusuke Mishina (National Institute of AIST), Kazuo Takaragi (National Institute of AIST)