Privacy and utility preserving sensor-data transformations

Malekzadeh, Mohammad; Clegg, R. E. S.; Cavallaro, Andrea; Haddadi, Hamed

doi:10.1016/j.pmcj.2020.101132

Cited by 30 publications

(18 citation statements)

References 38 publications

(55 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It aims at the trained systems that are deployed to offer Inference-as-a-Service [145]. Most methods in private inference are similar to those in private training, except for the Information-Theoretic privacy [138,137,139]. It is typically used to offer information-theoretic mathematical or empirical evidence of how these methods operate to improve privacy.…”

Section: Privacy Preservingmentioning

confidence: 99%

Socially Responsible AI Algorithms: Issues, Purposes, and Challenges

Cheng¹,

Varshney²,

Liu³

2021

Preprint

View full text Add to dashboard Cite

In the current era, people and society have grown increasingly reliant on Artificial Intelligence (AI) technologies. AI has the potential to drive us towards a future in which all of humanity flourishes. It also comes with substantial risks for oppression and calamity. Discussions about whether we should (re)trust AI have repeatedly emerged in recent years and in many quarters, including industry, academia, health care, services, and so on. Technologists and AI researchers have a responsibility to develop trustworthy AI systems. They have responded with great efforts of designing more responsible AI algorithms. However, existing technical solutions are narrow in scope and have been primarily directed towards algorithms for scoring or classification tasks, with an emphasis on fairness and unwanted bias. To build long-lasting trust between AI and human beings, we argue that the key is to think beyond algorithmic fairness and connect major aspects of AI that potentially cause AI's indifferent behavior. In this survey, we provide a systematic framework of Socially Responsible AI Algorithms that aims to examine the subjects of AI indifference and the need for socially responsible AI algorithms, define the objectives, and introduce the means by which we may achieve these objectives. We further discuss how to leverage this framework to improve societal well-being through protection, information, and prevention/mitigation.

show abstract

Section: Privacy Preservingmentioning

confidence: 99%

Socially Responsible AI Algorithms: Issues, Purposes, and Challenges

Cheng¹,

Varshney²,

Liu³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In particular, we show how to generate natural-looking adversarial images either by selectively modifying colors within chosen ranges that we perceive as natural or by enhancing details in the image. The references that will be covered in the tutorial are [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18].…”

Section: Context Motivation and Descriptionmentioning

confidence: 99%

Deep Learning for Privacy in Multimedia

Cavallaro

Malekzadeh

Shamsabadi

2020

Proceedings of the 28th ACM International Conference on Multimedia

Self Cite

View full text Add to dashboard Cite

We discuss the design and evaluation of machine learning algorithms that provide users with more control on the multimedia information they share. We introduce privacy threats for multimedia data and key features of privacy protection. We cover privacy threats and mitigating actions for images, videos, and motion-sensor data from mobile and wearable devices, and their protection from unwanted, automatic inferences. The tutorial offers theoretical explanations followed by examples with software developed by the presenters and distributed as open source. CCS CONCEPTS • Artificial intelligence • Machine learning • Human and societal aspects of security and privacy.

show abstract

“…The architecture of a DNN is mainly defined by the type and number of the layers, and the way these layers are connected to each other [41]. A commonly used DNN architecture to classify sensor data is composed of a convolutional neural network (CNN) followed by a feedforward neural network (FNN) or a recurrent neural network (RNN) [10], [11], [12], [13], [14], [15], [16], [17], [18]. A CNN is inherently adaptive to input data of variable dimensions and learns features relevant for the downstream task.…”

Section: Related Workmentioning

confidence: 99%

“…For example, static acceleration, that shows the magnitude and direction of to the earth's gravitational force, helps to recognize the wearer's posture; whereas dynamic acceleration, that shows changes in the motion velocity of the wearer, can be mapped to the wearer's different activities [8]. The time series generated by these motion sensors are processed over temporal windows and classified by deep neural networks (DNNs) [9], [10], [11], which process sensor data with pre-defined, fixed dimensions [11], [12], [13], [14], [15], [16], [17], [18], and cannot reliably handle, at inference time, dynamic situations (e.g. when the sampling rate changes or some sensors are dropped), which are important for energy preservation [19], [20], privacy protection [21], [22] and fault tolerance [23], [24].…”

Section: Introductionmentioning

confidence: 99%

DANA: Dimension-Adaptive Neural Architecture for Multivariate Sensor Data

Malekzadeh,

Clegg,

Cavallaro

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Current deep neural architectures for processing sensor data are mainly designed for data coming from a fixed set of sensors, with a fixed sampling rate. Changing the dimensions of the input data causes considerable accuracy loss, unnecessary computations, or application failures. To address this problem, we introduce a dimension-adaptive pooling (DAP) layer that makes deep architectures robust to temporal changes in sampling rate and in sensor availability. DAP operates on convolutional filter maps of variable dimensions and produces an input of fixed dimensions suitable for feedforward and recurrent layers. Building on this architectural improvement, we propose a dimension-adaptive training (DAT) procedure to generalize over the entire space of feasible data dimensions at the inference time. DAT comprises the random selection of dimensions during the forward passes and optimization with accumulated gradients of several backward passes. We then combine DAP and DAT to transform existing non-adaptive deep architectures into a Dimension-Adaptive Neural Architecture (DANA) without altering other architectural aspects. Our solution does not need up-sampling or imputation, thus reduces unnecessary computations at inference time. Experimental results on public datasets show that DANA prevents losses in classification accuracy of the state-of-the-art deep architectures, under dynamic sensor availability and varying sampling rates.

show abstract

Privacy and utility preserving sensor-data transformations

Cited by 30 publications

References 38 publications

Socially Responsible AI Algorithms: Issues, Purposes, and Challenges

Socially Responsible AI Algorithms: Issues, Purposes, and Challenges

Deep Learning for Privacy in Multimedia

DANA: Dimension-Adaptive Neural Architecture for Multivariate Sensor Data

Contact Info

Product

Resources

About