cv
Basics
| Name | Andrea Siposova |
| Role | Machine Learning Researcher |
| siposova.andrea@gmail.com | |
| Website | andreasiposova.github.io |
| Summary | Machine Learning researcher with 6+ years of experience, specializing in AI security and privacy-preserving machine learning. My experience includes designing and executing complex experiments, and deriving insights for improving model security and data confidentiality. Proven ability in project management and collaboration with cross-functional teams to advance research and development initiatives. Holding a Master’s degree in Data Science from Vienna University of Technology, with a thesis in the field of machine learning security. |
Work
-
2021 - 2024 -
2017 - 2017 -
2017 - 2019
Education
-
2019 - 2023 Vienna, Austria
MSc. (Dipl.Ing.)
TU Wien (Vienna University of Technology)
Data Science
- Diploma Thesis: Data Exfiltration Attacks and Defenses in Neural Networks:
- Simulated data exfiltration attacks on neural networks through intentional memorization in white-box and black-box scenarios.
- Conducted in-depth evaluation of attack effectiveness regarding various attack hyperparameters and evaluated attack vs. defense vs. model utility trade-offs.
- Designed and applied defense mechanisms, enhancing neural network security and robustness against exfiltration threats.
- Introduced a taxonomy of data exfiltration attacks in machine learning.
- Thesis & Defensio grade: 100%, Coursework grade average: 89%
-
2015 - 2018 Vienna, Austria
BSc.
WU Wien (Vienna University of Economics and Business)
Information Systems
- Data Science, Business Information Systems
Skills
| Programming & Data Science | |
| Python | |
| R |
| Libraries & Frameworks | |
| NumPy, pandas, scikit-learn, PyTorch, Tensorflow, ONNX, Hugging Face, LangChain, OpenAI, SciPy, matplotlib, seaborn |
| Development Tools | |
| Git, Docker | |
| Experiment tracking & Registry: Weights & Biases, MLFlow |
| Data | |
| Reational DBs, GraphDBs, VectorDB | |
| SQL, Cypher (neo4j), Qdrant, FAISS |
| Other | |
| Scientific writing: LaTeX | |
| OS: Linux | |
| MS Office |
Publications
-
2024 Datenexfiltration mit Hilfe von Modellen des maschinellen Lernens
OCG Journal. Austrian Computer Society. 01:20-21.
-
2023 Data Exfiltration Attacks and Defenses in Neural Networks
Diploma Thesis, TU Wien
-
2023 Supporting Landslide Disaster Risk Reduction Using Data-driven Methods
ERCIM NEWS-European Research Consortium for Informatics and Mathematics, 135:10-11.
-
2022 gAia: predicting landslides based on consolidated inventory data–bridging needs and limitations.
Disaster Research Days 2022. 43-45.
-
2020
Languages
| Slovak | |
| Native speaker |
| English | |
| Full proficiency |
| German | |
| Full proficiency |
| Czech | |
| Full proficiency |
| Hebrew | |
| Intermediate proficiency |
Interests
| Privacy-Preserving ML |
| Adversarial ML |
| AI Safety |
| Digital Rights Management / IP Protection |