Dhiria: unique technology

Scientific Research

The technological engine of Dhiria stems from the commitment to constantly and significantly advance scientific research. This means tackling and solving the challenges posed by the international scientific community. Thanks to our investments, Dhiria, the first in the world, has achieved significant results in privacy-preserving computation and is already able to bring this technological advancement to market.

More specifically, what our research has produced is the ability to process encrypted data using machine and deep learning algorithms in an “as-a-service” model: at every stage of processing, both the data and results remain encrypted, and at no point are they decrypted.

This remarkable achievement is the result of scientific advancements capable of combining new neural architectures and training algorithms that preserve user privacy with innovative machine and deep learning computation paradigms “as-a-service”.

DHIRIA is the first company in the world to offer machine and deep learning as-a-service on encrypted data.

Challenges

First in the world, Dhiria has tackled and solved three major scientific and technological challenges:

• Algorithmic challenge: Machine and deep learning models have been redesigned and developed to operate on encrypted data.

• Technological challenge: Privacy-preserving computation demands high memory and computational resources. Machine and deep learning algorithms have been re-engineered to reduce computational load, memory usage, and energy consumption, while ensuring processing accuracy.

• Architectural challenge: To bring efficient and effective machine and deep learning solutions capable of operating on encrypted data to market, we implemented “as-a-service” approaches where processing is carried out on high-performance systems in data centers or on the Cloud.

We contribute to advancing scientific research on an international scale to ensure the best technology for our clients.

Scientific Publications

The innovative research solutions and key results achieved by Dhiria in these areas have been published in leading international journals and conferences:

• A. Falcetta, M. Roveri, “EVAD: encrypted vibrational anomaly detection with homomorphic encryption”, Neural Computing and Applications, Springer 2024

• L. Colombo, A. Falcetta, M. Roveri, “TIFeD: a Tiny Integer-based Federated learning algorithm with Direct feedback alignment”, 2023 International Conference on AI-ML Systems, Bangalore, India, 2023.

• A. Falcetta, M. Pavan, S. Canali, V. Schiaffonati, M. Roveri, “To Personalize or Not To Personalize? Soft Personalization and the Ethics of ML for Health”, The 10th IEEE International Conference on Data Science and Advanced Analytics (DSAA), Thessaloniki, Greece, 2023.

• A. Falcetta, M. Roveri, “T4C: A Framework for Time-Series Clustering-as-a-Service”, CEUR Workshop Proceedings vol. 3252 – Workshop proceedings, 2023.

• F. Puoti, A. Falcetta, M. Roveri, D. Riva, D. Chiggiato, “GreenTea: Time-series Exploration as-a-Service for environmental science”, IEEE Conference on Artificial Intelligence (CAI), Santa Clara, CA, USA, 2023.

• M. Gambella, A. M. Roveri, “EDANAS: Adaptive Neural Architecture Search for Early Exit Neural Networks”, the International Joint Conference on Neural Networks (IEEE IJCNN), Gold Coast, Queensland, Australia, 2023.

• A. Falcetta, M. Roveri, “Privacy-preserving machine learning with homomorphic encryption: an introduction”, IEEE Computational Intelligence Magazine, August, 2022 [Q1 SCIMAGO – Computer Science].

• M. Gambella, A. Falcetta, M. Roveri, “CNAS: Constrained Neural Architecture Search”, Accepted at the 2022 IEEE International Conference on Systems, Man and Cybernetics (IEEE SMC 2022),Prague, 2022.

• A. Falcetta, M. Roveri, “TIMEX: an Automatic Framework for Time-Series Forecasting-as-a-Service”, Accepted at the 6th International Workshop on Automation in Machine Learning, held in conjunction with the SIGKDD Conference on Knowledge Discovery and Data Mining conference (KDD2022),Washington DC, 2022.

• A. Falcetta, M. Roveri, “Privacy-preserving time series prediction with temporal convolutional neural networks”, in Proc. 2022 International Joint Conference on Neural Network (IJCNN2022),Padova, Italy, 2022.

• S. Disabato, A. Falcetta, A. Mongelluzzo, M. Roveri, “A privacy-preserving distributed architecture for deep-learning-as-a-service”, in Proc. 2020 International Joint Conference on Neural Networks (IJCNN),Glasgow, 2020.

These publications certify the value and validity of Dhiria's solutions, as recognized by the international scientific community, in addressing the challenges of privacy-preserving computation.

Greater efficiency of algorithms and cloud infrastructures, designed to process encrypted data, means lower costs and higher performance for Dhiria's clients.

Dhiria: Competitive Advantages

One additional point deserves attention: the results obtained from Dhiria's research, designed to provide effective and efficient solutions in the field of privacy-preserving machine learning, also represent a valuable competency for Dhiria in plaintext computation.

Being able to process encrypted data introduces complexities that are one or even two orders of magnitude greater compared to equivalent plaintext processing, both in terms of processing time and the size of neural networks. This constantly drives us toward optimizing every component of our platform.

The innovative solutions developed by Dhiria in the algorithmic, technological, and architectural fields provide an incredible competitive advantage over the competition in bringing machine and deep learning as-a-service solutions to market that operate on plaintext, ensuring better performance and lower processing costs.