scholarly journals Protecting Private Inputs: Bounded Distortion Guarantees With Randomised Approximations

2020 ◽  
Vol 2020 (3) ◽  
pp. 284-303
Author(s):  
Patrick Ah-Fat ◽  
Michael Huth
Keyword(s):  
Experimental Work ◽  
Differential Privacy ◽  
Use Cases ◽  
Distortion Bounds ◽  
Low Probability ◽  
Bounded Distortion ◽  
Class Of Functions

AbstractComputing a function of some private inputs while maintaining the confidentiality of those inputs is an important problem, to which Differential Privacy and Secure Multi-party Computation can offer solutions under specific assumptions. Research in randomised algorithms aims at improving the privacy of such inputs by randomising the output of a computation while ensuring that large distortions of outputs occur with low probability. But use cases such as e-voting or auctions will not tolerate large distortions at all. Thus, we develop a framework for randomising the output of a privacypreserving computation, while guaranteeing that output distortions stay within a specified bound. We analyse the privacy gains of our approach and characterise them more precisely for our notion of sparse functions. We build randomisation algorithms, running in linearithmic time in the number of possible input values, for this class of functions and we prove that the computed randomisations maximise the inputs’ privacy. Experimental work demonstrates significant privacy gains when compared with existing approaches that guarantee distortion bounds, also for non-sparse functions.

scholarly journals Towards Understanding the Risks of Gradient Inversion in Federated Learning

2021 ◽  
Author(s):  
Ali Hatamizadeh ◽  
Hongxu Yin ◽  
Pavlo Molchanov ◽  
Andriy Myronenko ◽  
Wenqi Li ◽  
...  
Keyword(s):  
Neural Networks ◽  
Data Privacy ◽  
Deep Neural Networks ◽  
Differential Privacy ◽  
Use Cases ◽  
Training Data ◽  
Model Accuracy ◽  
Healthcare Applications ◽  
Raw Data ◽  
Collaborative Training

Abstract Federated learning (FL) allows the collaborative training of AI models without needing to share raw data. This capability makes it especially interesting for healthcare applications where patient and data privacy is of utmost concern. However, recent works on the inversion of deep neural networks from model gradients raised concerns about the security of FL in preventing the leakage of training data. In this work, we show that these attacks presented in the literature are impractical in real FL use-cases and provide a new baseline attack that works for more realistic scenarios where the clients’ training involves updating the Batch Normalization (BN) statistics. Furthermore, we present new ways to measure and visualize potential data leakage in FL. Our work is a step towards establishing reproducible methods of measuring data leakage in FL and could help determine the optimal tradeoffs between privacy-preserving techniques, such as differential privacy, and model accuracy based on quantifiable metrics.

scholarly journals Towards Understanding the Risks of Gradient Inversion in Federated Learning

2021 ◽  
Author(s):  
Ali Hatamizadeh ◽  
Hongxu Yin ◽  
Pavlo Molchanov ◽  
Andriy Myronenko ◽  
Wenqi Li ◽  
...  
Keyword(s):  
Neural Networks ◽  
Data Privacy ◽  
Deep Neural Networks ◽  
Differential Privacy ◽  
Use Cases ◽  
Training Data ◽  
Model Accuracy ◽  
Healthcare Applications ◽  
Raw Data ◽  
Collaborative Training

Abstract Federated learning (FL) allows the collaborative training of AI models without needing to share raw data. This capability makes it especially interesting for healthcare applications where patient and data privacy is of utmost concern. However, recent works on the inversion of deep neural networks from model gradients raised concerns about the security of FL in preventing the leakage of training data. In this work, we show that these attacks presented in the literature are impractical in real FL use-cases and provide a new baseline attack that works for more realistic scenarios where the clients’ training involves updating the Batch Normalization (BN) statistics. Furthermore, we present new ways to measure and visualize potential data leakage in FL. Our work is a step towards establishing reproducible methods of measuring data leakage in FL and could help determine the optimal tradeoffs between privacy-preserving techniques, such as differential privacy, and model accuracy based on quantifiable metrics.

scholarly journals Study on the Technical Evaluation of De-Identification Procedures for Personal Data in the Automotive Sector

2021 ◽  
Author(s):  
Kai Rannenberg ◽  
Sebastian Pape ◽  
Frédéric Tronnier ◽  
Sascha Löbner
Keyword(s):  
Data Privacy ◽  
Differential Privacy ◽  
Homomorphic Encryption ◽  
Personal Data ◽  
Legal Aspects ◽  
Use Cases ◽  
Use Case ◽  
User Privacy ◽  
External Data ◽  
Identification Techniques

The aim of this study was to identify and evaluate different de-identification techniques that may be used in several mobility-related use cases. To do so, four use cases have been defined in accordance with a project partner that focused on the legal aspects of this project, as well as with the VDA/FAT working group. Each use case aims to create different legal and technical issues with regards to the data and information that are to be gathered, used and transferred in the specific scenario. Use cases should therefore differ in the type and frequency of data that is gathered as well as the level of privacy and the speed of computation that is needed for the data. Upon identifying use cases, a systematic literature review has been performed to identify suitable de-identification techniques to provide data privacy. Additionally, external databases have been considered as data that is expected to be anonymous might be reidentified through the combination of existing data with such external data. For each case, requirements and possible attack scenarios were created to illustrate where exactly privacy-related issues could occur and how exactly such issues could impact data subjects, data processors or data controllers. Suitable de-identification techniques should be able to withstand these attack scenarios. Based on a series of additional criteria, de-identification techniques are then analyzed for each use case. Possible solutions are then discussed individually in chapters 6.1 - 6.2. It is evident that no one-size-fits-all approach to protect privacy in the mobility domain exists. While all techniques that are analyzed in detail in this report, e.g., homomorphic encryption, differential privacy, secure multiparty computation and federated learning, are able to successfully protect user privacy in certain instances, their overall effectiveness differs depending on the specifics of each use case.

scholarly journals On Uniformly Bazilevic and Related Functions

2012 ◽  
Vol 2012 ◽  
pp. 1-15 ◽  
Author(s):  
Khalida Inayat Noor
Keyword(s):  
Open Unit ◽  
Arc Length ◽  
Open Unit Disc ◽  
New Class ◽  
Univalence Criterion ◽  
Distortion Bounds ◽  
Special Cases ◽  
Bazilevic Functions ◽  
Class Of Functions ◽  
Inclusion Results

We introduce a new class of functions analytic in the open unit disc, which contains the class of Bazilevic functions and also generalizes the concept of uniform convexity. We establish univalence criterion for the functions in this class and investigate rate of growth of coefficients, arc length problem, inclusion results, and distortion bounds. Some interesting results are derived as special cases.

scholarly journals Inequalities of harmonic univalent functions with connections of hypergeometric functions

2015 ◽  
Vol 13 (1) ◽  
Author(s):  
Janusz Sokół ◽  
Rabha W. Ibrahim ◽  
M. Z. Ahmad ◽  
Hiba F. Al-Janaby
Keyword(s):  
Unit Disk ◽  
Hypergeometric Functions ◽  
Univalent Functions ◽  
Extreme Points ◽  
Open Unit ◽  
Open Unit Disk ◽  
Distortion Bounds ◽  
Class Of Functions

AbstractLet SH be the class of functions f = h+g that are harmonic univalent and sense-preserving in the open unit disk U = { z : |z| < 1} for which f (0) = f'(0)-1=0. In this paper, we introduce and study a subclass H( α, β) of the class SH and the subclass NH( α, β) with negative coefficients. We obtain basic results involving sufficient coefficient conditions for a function in the subclass H( α, β) and we show that these conditions are also necessary for negative coefficients, distortion bounds, extreme points, convolution and convex combinations. In this paper an attempt has also been made to discuss some results that uncover some of the connections of hypergeometric functions with a subclass of harmonic univalent functions.

scholarly journals Private Reverse Top-k Algorithms Applied on Public Data of COVID-19 in the State of Ceará

2021 ◽  
Vol 12 (5) ◽  
Author(s):  
Maria de Lourdes M. Silva ◽  
Iago C. Chaves ◽  
Javam C. Machado
Keyword(s):  
Differential Privacy ◽  
Personal Data ◽  
The State ◽  
Experimental Results ◽  
Public Data ◽  
Low Probability ◽  
High Degree ◽  
Degree Of Similarity

In this article we propose a differentially private reverse top-k query. Our strategy allows obtaining the less frequent data according to a search criteria, with a high guarantee of privacy of the individuals who contributed with personal data in the original database. We apply our strategy on public data for COVID-19 in the State of Ceará using two different queries. Our experimental results show that the result of the proposed top-k query returns a high degree of similarity to the result of a conventional top-k query, when the chosen budget is suitable, providing useful results for researchers, while ensuring a low probability of re-identification of individuals arising from the properties of differential privacy.

Research into the Effect of Photometric Flicker Event on the Perception of Office Workers

2019 ◽  
pp. 22-27
Author(s):  
Cenk Yavuz ◽  
Ceyda Aksoy Tırmıkç ◽  
Burcu Çarklı Yavuz
Keyword(s):  
Experimental Work ◽  
Psychological Problems ◽  
Office Workers ◽  
Questionnaire Study ◽  
Fast Growing ◽  
Lighting Conditions ◽  
Lighting Systems ◽  
Enormous Number

Today the number of office workers has reached to an enormous number due to the fast-growing technology. Most of these office workers spend long hours in enclosed spaces with little/no daylight penetration. The lack of daylight causes physiological and psychological problems with the workers. At this point lighting systems become prominent as the source and the solution of the problem. Photometric flicker event which arises in the lighting systems can sometimes become visible and brings a lot of issues with it. In this paper, an experimental work has been done to investigate the effect of flicker. For this purpose, the flicker values of 3 different experiment rooms for different lighting conditions and scenarios have been measured and a questionnaire study has been carried out in the experiment rooms with 30 participants. In conclusion, the effect of the flicker event on the volunteers have been classified and some methods have been proposed not to experience flicker effects.

CLASS OF BOOLEAN FUNCTIONS CONSTRUCTED USING SIGNIFICANT BITS OF LINEAR RECURRENCES OVER THE RING ℤ2n

2019 ◽  
Vol 6 (2) ◽  
pp. 90-94
Author(s):  
Hernandez Piloto Daniel Humberto
Keyword(s):  
Linear Function ◽  
Boolean Functions ◽  
Hamming Distance ◽  
Linear Recurrences ◽  
Maximum Period ◽  
Non Linear ◽  
The Given ◽  
Class Of Functions

In this work a class of functions is studied, which are built with the help of significant bits sequences on the ring ℤ2n. This class is built with use of a function ψ: ℤ2n → ℤ2. In public literature there are works in which ψ is a linear function. Here we will use a non-linear ψ function for this set. It is known that the period of a polynomial F in the ring ℤ2n is equal to T(mod 2)2α, where α∈ , n01- . The polynomials for which it is true that T(F) = T(F mod 2), in other words α = 0, are called marked polynomials. For our class we are going to use a polynomial with a maximum period as the characteristic polyomial. In the present work we show the bounds of the given class: non-linearity, the weight of the functions, the Hamming distance between functions. The Hamming distance between these functions and functions of other known classes is also given.

THE ROLE OF LITERACY METHODS IN MASTERING OF WRITTEN LANGUAGE BY STUDENTS WITH DEVELOPMENTAL DYSLEXIA

2019 ◽  
Vol 7 (1) ◽  
pp. 10-18
Author(s):  
Tsvetanka Tsenova
Keyword(s):  
Experimental Work ◽  
Developmental Dyslexia ◽  
Reading Skills ◽  
Writing Skills ◽  
Written Language ◽  
Reading And Writing ◽  
Empirical Verification ◽  
Literacy Methods ◽  
The Relationship

This article focuses on the relationship between literacy methods applied at school and the emergence of serious difficulties in mastering reading and writing skills that shape the developmental dyslexia. The problem was analyzed theoretically and subjected to empirical verification. Experimental work was presented which aims to study the phonological and global reading skills of 4- th grade students with and without dyslexia. Better global reading skills have been demonstrated in all tested children, and this is much more pronounced in those with dyslexia than their peers without disorders. Hence, the need to develop a special, corrective methodology for literacy of students with developmental dyslexia consistent with their psychopathological characteristics.

Sign in / Sign up

Export Citation Format

Share Document