Andreea B. Alexandru

[jump to Publications]

Researcher

My main focus is in the privacy and security of dynamical systems. Many of the structures we see around us are dynamical systems; from the time series coming from medical monitoring sensors to the energy consumption of our homes. Moreover, abstract concepts can be modeled as dynamical systems, for example: iterative optimization algorithms like the gradient descent algorithm, as well as the processes of training and evaluating neural networks (ideas from control theory can help making decisions and controlling variables of interest in all these areas). Nevertheless, concealing dynamical data brings extra challenges, such as dealing with the dependencies between data at different time steps, maintaining privacy at consecutive iterations and accumulation of noise in the result.

Photo

My main research involves designing privacy-preserving control and optimization algorithms that make use of homomorphic encryption schemes and secure multi-party computation schemes. The idea behind these concepts is to perform computations directly on encrypted data, such that leakage of private information to the computing entity is minimized. So far, the main applications of my research lie in control theory and machine learning areas, as well as in general networked systems. A rough partition of my research projects consists of:

  • private cloud-based quadratic optimization from distributed private data;
  • linear and nonlinear cloud-based control on encrypted data;
  • private data-driven cloud-based control;
  • oblivious distributed weighted sum aggregation.

I am very interested in fully homomorphic encryption and in expanding and efficientizing its capabilities. In the summer of 2019, I worked at Duality Technologies, a start-up founded by academics in the field of cryptography, that centers on developing software for real-world encrypted computing applications. I focused on optimizing various encrypted capabilities in a proprietary version of PALISADE, a library for fully homomorphic encryption.

Another project related to privacy I worked on exploited the system's model to achieve motion planning with secrecy guarantees.

Previously, I worked on Decentralized Estimation for limited communication between sensors, exploiting graph theory and structural systems theory and investigated network sparsification based on controllability and observability measures.

My vision for the future involves research in three major areas:

  • continuing to enable trustworthy machine learning and control, by incorporating zero-knowledge verification and defenses against malicious adversaries;
  • developing lightweight privacy solutions for low-power distributed devices, vital to IoT;
  • contributing to a more practical fully homomorphic encryption implementation, such that it becomes widely available for data-heavy industries.
These goals require fundamental research for developing privacy-preserving solutions adapted to structured dynamical data, investigating novel techniques to achieve numerically stable and efficient algorithms and combining formal verification with trusted hardware to ensure safety and security.

Publications

Preprints

  • Alexandru A. B., and Pappas G. J., Private Weighted Sum Aggregation, 2020. arXiv
  • Alexandru A. B., Tsiamis A. and Pappas G. J., Data-driven Control on Encrypted Data, 2020. arXiv

Journals and book chapters

  • Schulze Darup M., Alexandru A. B., Quevedo D. E. and Pappas G. J., Encrypted control for networked systems -- An illustrative introduction and current challenges, to appear in the IEEE Control Systems, 2021. arXiv
  • Alexandru A. B. and Pappas G. J., Secure Multi-party Computation for Cloud-Based Control, in "Privacy in Dynamical Systems", pp. 179-207, 2020, Springer, Singapore. arXiv.
  • Alexandru A. B., Gatsis K., Shoukry Y., Seshia S. A., Tabuada P. and Pappas, G. J., Cloud-based Quadratic Optimization with Partially Homomorphic Encryption, IEEE Transactions on Automatic Control, 2020. arXiv, GitHub.

Conferences

  • Alexandru A. B., Tsiamis A. and Pappas G. J., Towards Private Data-driven Control, in Proceedings of the 59th Conference on Decision and Control (CDC), pp. 5449-5456, 2020. 2020, IEEE.
  • Alexandru A. B. and Pappas G. J., Private Weighted Sum Aggregation for Distributed Control Systems, in Proceedings of the 21st International Federation of Automatic Control (IFAC) World Congress, 2020. paper
  • Alexandru A. B., Schulze Darup M. and Pappas G. J., Encrypted cooperative control revisited, in Proceedings of the 58th Conference on Decision and Control (CDC), pp. 7196-7202, 2019, IEEE.
  • Tsiamis A., Alexandru A. B. and Pappas G. J., Motion Planning with Secrecy, in Proceedings of the American Control Conference (ACC), pp. 784-791, 2019, IEEE. Finalist for best student paper award.
  • Alexandru A. B. and Pappas G. J., Encrypted LQG using Labeled Homomorphic Encryption, in Proceedings of 10th ACM/IEEE International Conference on Cyber-Physical Systems (ICCPS), pp. 129-140, 2019. Finalist for best paper award. GitHub, correction.
  • Alexandru A. B., Morari M. and Pappas G. J., Cloud-based MPC with Encrypted Data, in Proceedings of the 57th Conference on Decision and Control (CDC), pp. 5014-5019 2018, IEEE. arXiv extended version, GitHub.
  • Alexandru A.B., Pequito S., Jadbabaie A. and Pappas G.J., 2017, On the Limited Communication Analysis and Design for Decentralized Estimation in Proceedings of the 56th Conference on Decision and Control (CDC), pp. 1713-1718, IEEE. arXiv extended version.
  • Alexandru A.B., Gatsis K. and Pappas G.J., 2017, Privacy preserving Cloud-based Quadratic Optimization in Proceedings of the 55th Annual Allerton Conference on Communication, Control, and Computing, pp. 1168-1175, IEEE.
  • Alexandru A.B., Pequito S., Jadbabaie A. and Pappas G.J., 2016, Decentralized observability with limited communication between sensors in Proceedings of the 55th Conference on Decision and Control (CDC), pp. 885-890, IEEE. arXiv extended version.
  • Alexandru A.B., Lup, S., Dita B., 2013, GDS2M: Preprocessing Tool for MEMS Devices in Proceedings of the 8th International Symposium on Advanced Topics in Electrical Engineering (ATEE), pp. 1-4, IEEE. Third place in best student paper competition.

Talks

Apart from presenting my conference papers at their respective venues, I have also given the following talks:

  • Cloud-based MPC with Encrypted Data, ESE Department PhD Colloquium, Oct. 2018, University of Pennsylvania.
  • Privacy Preserving Cloud-based Quadratic Optimization, ESE Department PhD Colloquium, Oct. 2017, University of Pennsylvania.
  • Secure Cloud-outsourced Optimization Problems through Homomorphic Encryption, Aug. 2017, Intel-NSF Center on Cyber Physical System Security.

Posters

  • Privacy for Cyber-Physical Systems, Oct. 2019, EECS Rising Stars at UIUC.
  • Private Cooperative Control, Oct. 2019, Grace Hopper Celebration.
  • Privacy for Cyber-Physical Systems, Mar. 2019, ECEDHA Annual Conference, iREDEFINE Workshop.
  • Privacy preserving Cloud-based Quadratic Optimization, Mar. 2018, 5th Annual Women in Cybersecurity Conference.
  • Secure Cloud-outsourced Optimization Problems through Homomorphic Encryption, Aug. 2017, Intel-NSF Center on Cyber Physical System Security.