When the signal is in the noise: The limits of Diffix's sticky noise

by   Andrea Gadotti, et al.

Finding a balance between privacy and utility, allowing researchers and businesses to use data for good while protecting people's privacy, is one of the biggest challenge we face today. A large body of research has shown the limits of the traditional anonymization (or de-identification) model prompting the use of question and answer or query-based systems. Diffix is a query-based system developed by Aircloak using the concept of "sticky noise" to protect people's privacy. We here present an attack on Diffix that exploits the structure of its sticky noise to infer private attributes of people in the dataset. We believe this vulnerability to be serious, allowing us to accurately infer private information of users with little background knowledge. While we share Diffix's creators view that we need to take a fresh look at building practical privacy-preserving systems, we believe this requires a layered security approach and fully open tools and discussions. Patented and proprietary code is unlikely to be sufficient to truly help us find a balance between the great potential of data and the basic human right of privacy.


page 1

page 2

page 3

page 4


An Efficient and Scalable Privacy Preserving Algorithm for Big Data and Data Streams

A vast amount of valuable data is produced and is becoming available for...

Learning to Collaborate for User-Controlled Privacy

It is becoming increasingly clear that users should own and control thei...

E-DPNCT: An Enhanced Attack Resilient Differential Privacy Model For Smart Grids Using Split Noise Cancellation

High frequency reporting of energy utilization data in smart grids can b...

I Am Not What I Write: Privacy Preserving Text Representation Learning

Online users generate tremendous amounts of textual information by parti...

Survey and Open Problems in Privacy Preserving Knowledge Graph: Merging, Query, Representation, Completion and Applications

Knowledge Graph (KG) has attracted more and more companies' attention fo...

Does a Face Mask Protect my Privacy?: Deep Learning to Predict Protected Attributes from Masked Face Images

Contactless and efficient systems are implemented rapidly to advocate pr...

To share or not to share: What risks would laypeople accept to give sensitive data to differentially-private NLP systems?

Although the NLP community has adopted central differential privacy as a...

Please sign up or login with your details

Forgot password? Click here to reset