Secure decentralized peer-to-peer training of deep neural networks based on distributed ledger technology
- PDF / 1,110,235 Bytes
- 15 Pages / 439.37 x 666.142 pts Page_size
- 81 Downloads / 183 Views
Secure decentralized peer‑to‑peer training of deep neural networks based on distributed ledger technology Amin Fadaeddini1 · Babak Majidi1 · Mohammad Eshghi2
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The accuracy and performance of deep neural network models become important issues as the applications of deep learning increase. For example, the navigation system of autonomous self-driving vehicles requires very accurate deep learning models. If a self-driving car fails to detect a pedestrian in bad weather, the result can be devastating. If we can increase the model accuracy by increasing the training data, the probability of avoiding such scenarios increases significantly. However, the problem of privacy for consumers and lack of enthusiasm for sharing their personal data, e.g., the recordings of their self-driving car, is an obstacle for using this valuable data. In Blockchain technology, many entities which cannot trust each other in normal conditions can join together to achieve a mutual goal. In this paper, a secure decentralized peer-to-peer framework for training the deep neural network models based on the distributed ledger technology in Blockchain ecosystem is proposed. The proposed framework anonymizes the identity of data providers and therefore can be used as an incentive for consumers to share their private data for training deep learning models. The proposed framework uses the Stellar Blockchain infrastructure for secure decentralized training of the deep models. A deep learning coin is proposed for Blockchain compensation. Keywords Deep learning · Privacy-preserving · Blockchain · Autonomous selfdriving car
* Babak Majidi [email protected] 1
Department of Computer Engineering, Khatam University, North Shiraz Street, Tehran, Iran
2
Department of Computer Engineering, Shahid Beheshti University, Tehran, Iran
13
Vol.:(0123456789)
A. Fadaeddini et al.
1 Introduction The promise of deep learning models is to find the complex and nonlinear patterns in the datasets. This promise has been proven by acceptable results in domains like object recognition, synthesizing samples, image retrieval and many other applications [1–6]. However, in many of these applications, security and privacy of the training data for deep models are the vital issue. These aforementioned advances in deep learning are strongly dependent on the presence of big datasets. Without this large datasets, trained deep models will fail to achieve acceptable accuracy. Unfortunately, this valuable information cannot be accessed without considerable cost and effort. A large percentage of datasets are in possession of larger companies and are only available at a high cost. This cost makes it almost impossible for startups and researchers to train high accuracy deep models. Along with strict legal regulations which will restrict sharing of end user’s data, there are also privacy concerns for ordinary individuals to share their personal data due to possibility of these private data to become av
Data Loading...
