Publications
158 results found
Aloufi R, Haddadi H, Boyle D, 2019, Emotionless: privacy-preserving speech analysis for voice assistants, Publisher: arXiv
Voice-enabled interactions provide more human-like experiences in manypopular IoT systems. Cloud-based speech analysis services extract usefulinformation from voice input using speech recognition techniques. The voicesignal is a rich resource that discloses several possible states of a speaker,such as emotional state, confidence and stress levels, physical condition, age,gender, and personal traits. Service providers can build a very accurateprofile of a user's demographic category, personal preferences, and maycompromise privacy. To address this problem, a privacy-preserving intermediatelayer between users and cloud services is proposed to sanitize the voice input.It aims to maintain utility while preserving user privacy. It achieves this bycollecting real time speech data and analyzes the signal to ensure privacyprotection prior to sharing of this data with services providers. Precisely,the sensitive representations are extracted from the raw signal by usingtransformation functions and then wrapped it via voice conversion technology.Experimental evaluation based on emotion recognition to assess the efficacy ofthe proposed method shows that identification of sensitive emotional state ofthe speaker is reduced by ~96 %.
Mo F, Shamsabadi AS, Katevas K, et al., 2019, Towards Characterizing and Limiting Information Exposure in DNN Layers
Pre-trained Deep Neural Network (DNN) models are increasingly used insmartphones and other user devices to enable prediction services, leading topotential disclosures of (sensitive) information from training data capturedinside these models. Based on the concept of generalization error, we propose aframework to measure the amount of sensitive information memorized in eachlayer of a DNN. Our results show that, when considered individually, the lastlayers encode a larger amount of information from the training data compared tothe first layers. We find that, while the neuron of convolutional layers canexpose more (sensitive) information than that of fully connected layers, thesame DNN architecture trained with different datasets has similar exposure perlayer. We evaluate an architecture to protect the most sensitive layers withinthe memory limits of Trusted Execution Environment (TEE) against potentialwhite-box membership inference attacks without the significant computationaloverhead.
Malekzadeh M, Clegg RG, Cavallaro A, et al., 2019, Mobile sensor data anonymization, ACM/IEEE International Conference on Internet of Things Design and Implementation (IoTDI 2019), Publisher: ACM, Pages: 49-58
Data from motion sensors such as accelerometers and gyroscopes embedded inour devices can reveal secondary undesired, private information about ouractivities. This information can be used for malicious purposes such as useridentification by application developers. To address this problem, we propose adata transformation mechanism that enables a device to share data for specificapplications (e.g.~monitoring their daily activities) without revealing privateuser information (e.g.~ user identity). We formulate this anonymization processbased on an information theoretic approach and propose a new multi-objectiveloss function for training convolutional auto-encoders~(CAEs) to provide apractical approximation to our anonymization problem. This effective lossfunction forces the transformed data to minimize the information about theuser's identity, as well as the data distortion to preserveapplication-specific utility. Our training process regulates the encoder todisregard user-identifiable patterns and tunes the decoder to shape the finaloutput independently of users in the training set. Then, a trained CAE can bedeployed on a user's mobile device to anonymize sensor data before sharing withan app, even for users who are not included in the training dataset. Theresults, on a dataset of 24 users for activity recognition, show a promisingtrade-off on transformed data between utility and privacy, with an accuracy foractivity recognition over 92%, while reducing the chance of identifying a userto less than 7%.
Moore J, Arcia-Moret A, Yadav P, et al., 2019, Zest: REST over ZeroMQ, 2019 IEEE INTERNATIONAL CONFERENCE ON PERVASIVE COMPUTING AND COMMUNICATIONS WORKSHOPS (PERCOM WORKSHOPS), Pages: 1015-1019, ISSN: 2474-2503
Moore J, Arcia-Moret A, Yadav P, et al., 2019, Zest: REST over ZeroMQ, Pages: 1015-1019
In this paper we introduce, Zest (REST over ZeroMQ), a middleware technology in support of an Internet of Things (IoT). Our work is influenced by the Constrained Application Protocol (CoAP) but emphasises systems that can support fine-grained access control to both resources and audit information, and can provide features such as asynchronous communication patterns between nodes. We achieve this by using a hybrid approach that combines a RESTful architecture with a variant of a publisher/subscriber topology that has enhanced routing support. The primary motivation for Zest is to provide inter-component communications in the Databox, but it is applicable in other contexts where tight control needs to be maintained over permitted communication patterns.
Zhan Y, Haddadi H, 2019, Towards Automating Smart Homes: Contextual and Temporal Dynamics of Activity Prediction, ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) / ACM International Symposium on Wearable Computers (ISWC), Publisher: ASSOC COMPUTING MACHINERY, Pages: 413-417
Zhan Y, Haddadi H, 2019, Activity Prediction for Improving Well-Being of Both The Elderly and Caregivers, ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) / ACM International Symposium on Wearable Computers (ISWC), Publisher: ASSOC COMPUTING MACHINERY, Pages: 1214-1217
Zhan Y, Haddadi H, 2019, Activity Prediction for Mapping Contextual-Temporal Dynamics, ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) / ACM International Symposium on Wearable Computers (ISWC), Publisher: ASSOC COMPUTING MACHINERY, Pages: 246-249
Aloufi R, Haddadi H, Boyle D, 2019, Poster Abstract: Privacy Preserving Speech Analysis using Emotion Filtering at the Edge, 17th ACM Conference on Embedded Networked Sensor Systems (SenSys), Publisher: ASSOC COMPUTING MACHINERY, Pages: 426-427
Popescu DA, Safronov V, Yadav P, et al., 2019, Poster Abstract: "Sensing" the IoT Network: Ethical Capture of Domestic IoT Network Traffic, 17th ACM Conference on Embedded Networked Sensor Systems (SenSys), Publisher: ASSOC COMPUTING MACHINERY, Pages: 406-407
Varvello M, Katevas K, Hang W, et al., 2019, Demo Abstract: BatteryLab, A Distributed Power Monitoring Platform For Mobile Devices, 17th ACM Conference on Embedded Networked Sensor Systems (SenSys), Publisher: ASSOC COMPUTING MACHINERY, Pages: 386-387
Mo F, Shamsabadi AS, Katevas K, et al., 2019, Poster: Towards Characterizing and Limiting Information Exposure in DNN Layers, ACM SIGSAC Conference on Computer and Communications Security (CCS), Publisher: ASSOC COMPUTING MACHINERY, Pages: 2653-2655
Varvello M, Katevas K, Plesa M, et al., 2019, BatteryLab, A Distributed Power Monitoring Platform For Mobile Devices, 18th ACM Workshop on Hot Topics in Networks (HotNets), Publisher: ASSOC COMPUTING MACHINERY, Pages: 101-108
- Author Web Link
- Cite
- Citations: 2
Osia SA, Rassouli B, Haddadi H, et al., 2019, Privacy Against Brute-Force Inference Attacks, Publisher: IEEE
- Author Web Link
- Cite
- Citations: 4
Zhang C, Patras P, Haddadi H, 2019, Deep Learning in Mobile and Wireless Networking: A Survey, IEEE COMMUNICATIONS SURVEYS AND TUTORIALS, Vol: 21, Pages: 2224-2287
- Author Web Link
- Cite
- Citations: 713
Katevas K, Hansel K, Clegg R, et al., 2019, Finding Dory in the Crowd: Detecting Social Interactions using Multi-Modal Mobile Sensing, SENSYS-ML'19: PROCEEDINGS OF THE FIRST WORKSHOP ON MACHINE LEARNING ON EDGE IN SENSOR SYSTEMS, Pages: 37-42
- Author Web Link
- Cite
- Citations: 5
Haddadi H, Crowcroft J, 2018, Welcome Message from the ICNP 2018 General Chairs, Proceedings - International Conference on Network Protocols, ICNP, Vol: 2018-September, ISSN: 1092-1648
Shamsabadi AS, Haddadi H, Cavallaro A, 2018, Distributed One-Class Learning, Pages: 4123-4127, ISSN: 1522-4880
We propose a cloud-based filter trained to block third parties from uploading privacy-sensitive images of others to online social media. The proposed filter uses Distributed One-Class Learning, which decomposes the cloud-based filter into multiple one-class classifiers. Each one-class classifier captures the properties of a class of privacy-sensitive images with an autoencoder. The multi-class filter is then reconstructed by combining the parameters of the one-class autoen-coders. The training takes place on edge devices (e.g. smartphones) and therefore users do not need to upload their private and/or sensitive images to the cloud. A major advantage of the proposed filter over existing distributed learning approaches is that users cannot access, even indirectly, the parameters of other users. Moreover, the filter can cope with the imbalanced and complex distribution of the image content and the independent probability of addition of new users. We evaluate the performance of the proposed distributed filter using the exemplar task of blocking a user from sharing privacy-sensitive images of other users. In particular, we validate the behavior of the proposed multi-class filter with non-privacy-sensitive images, the accuracy when the number of classes increases, and the robustness to attacks when an adversary user has access to privacy-sensitive images of other users.
Hänsel K, Katevas K, Orgs G, et al., 2018, The potential of wearable technology for monitoring social interactions based on interpersonal synchrony, Pages: 45-47
Sensing data from wearables have been extensively evaluated for fitness tracking, health monitoring or rehabilitation of individuals. However, we believe that wearable sensing can go beyond the individual and offer insights into social dynamics and interactions with other users by considering multi-user data. In this work, we present a new approach to using wrist-worn wearables for social monitoring and the detection of social interaction features based on interpersonal synchrony - An approach transferable to smartwatches and fitness trackers. We build up on related work in the field of psychology and present a study where we collected wearable sensing data during a social event with 24 participants. Our preliminary results indicate differences in wearable sensing data during a social interaction between two people.
Servia-Rodriguez S, Wang L, Zhao JR, et al., 2018, Privacy-preserving personal model training, Proceedings - ACM/IEEE International Conference on Internet of Things Design and Implementation, IoTDI 2018, Pages: 153-164
© 2018 IEEE. Many current Internet services rely on inferences from models trained on user data. Commonly, both the training and inference tasks are carried out using cloud resources fed by personal data collected at scale from users. Holding and using such large collections of personal data in the cloud creates privacy risks to the data subjects, but is currently required for users to benefit from such services. We explore how to provide for model training and inference in a system where computation is pushed to the data in preference to moving data to the cloud, obviating many current privacy risks. Specifically, we take an initial model learnt from a small set of users and retrain it locally using data from a single user. We evaluate on two tasks: one supervised learning task, using a neural network to recognise users' current activity from accelerometer traces; and one unsupervised learning task, identifying topics in a large set of documents. In both cases the accuracy is improved. We also analyse the robustness of our approach against adversarial attacks, as well as its feasibility by presenting a performance evaluation on a representative resource-constrained device (a Raspberry Pi).
Osia SA, Shamsabadi AS, Taheri A, et al., 2018, Private and scalable personal data analytics using hybrid edge-to-cloud deep learning, Computer, Vol: 51, Pages: 42-49, ISSN: 0018-9162
Although the ability to collect, collate, and analyze the vast amount of data generated from cyber-physical systems and Internet of Things devices can be beneficial to both users and industry, this process has led to a number of challenges, including privacy and scalability issues. The authors present a hybrid framework where user-centered edge devices and resources can complement the cloud for providing privacy-aware, accurate, and efficient analytics.
Malekzadeh M, Clegg RG, Cavallaro A, et al., 2018, Protecting sensory data against sensitive inferences, Workshop on Privacy by Design in Distributed Systems 2018
There is growing concern about how personal data are used when users grant applications direct access to the sensors in their mobile devices. For example,time-series data generated by motion sensors reflect directly users' activitiesand indirectly their personalities. It is therefore important to designprivacy-preserving data analysis methods that can run on mobile devices. Inthis paper, we propose a feature learning architecture that can be deployed indistributed environments to provide flexible and negotiable privacy-preservingdata transmission. It should be flexible because the internal architecture ofeach component can be independently changed according to users or serviceproviders needs. It is negotiable because expected privacy and utility can benegotiated based on the requirements of the data subject and underlyingapplication. For the specific use-case of activity recognition, we conductedexperiments on two real-world datasets of smartphone's motion sensors, one ofthem is collected by the authors and will be publicly available by this paperfor the first time. Results indicate the proposed framework establishes a goodtrade-off between application's utility and data subjects' privacy. We showthat it maintains the usefulness of the transformed data for activityrecognition (with around an average loss of three percentage points) whilealmost eliminating the possibility of gender classification (from more than90\% to around 50\%, the target random guess). These results also haveimplication for moving from the current binary setting of granting permissionto mobile apps or not, toward a situation where users can grant eachapplication permission over a limited range of inferences according to theprovided services.
Hänsel K, Poguntke R, Haddadi H, et al., 2018, What to put on the user: Sensing technologies for studies and physiology aware systems, ACM Conference on Human Factors in Computing Systems (ACM CHI’18), Publisher: ACM
Fitness trackers not just provide easy means to acquire physiological data in real-world environments due to affordable sensing technologies, they further offer opportunities for physiology-aware applications and studies in HCI; however, their performance is not well understood. In this paper, we report findings on the quality of 3 sensing technologies: PPG-based wrist trackers (Apple Watch, Microsoft Band 2), an ECG-belt (Polar H7) and reference device with stick-on ECG electrodes (Nexus 10). We collected physiological (heart rate, electrodermal activity, skin temperature) and subjective data from 21 participants performing combinations of physical activity and stressful tasks. Our empirical research indicates that wrist devices provide a good sensing performance in stationary settings. However, they lack accuracy when participants are mobile or if tasks require physical activity. Based on our findings, we suggest a textitDesign Space for Wearables in Research Settings and reflected on the appropriateness of the investigated technologies in research contexts.
Chamberlain A, Crabtree A, Haddadi H, et al., 2018, Special theme on privacy and the Internet of things, PERSONAL AND UBIQUITOUS COMPUTING, Vol: 22, Pages: 289-292, ISSN: 1617-4909
Crabtree A, Lodge T, Colley J, et al., 2018, Building accountability into the Internet of Things: the IoT Databox model, Journal of Reliable Intelligent Environments, Vol: 4, Pages: 39-55, ISSN: 2199-4668
This paper outlines the IoT Databox model as a means of making the Internet of Things (IoT) accountable to individuals. Accountability is a key to building consumer trust and is mandated by the European Union’s general data protection regulation (GDPR). We focus here on the ‘external’ data subject accountability requirement specified by GDPR and how meeting this requirement turns on surfacing the invisible actions and interactions of connected devices and the social arrangements in which they are embedded. The IoT Databox model is proposed as an in principle means of enabling accountability and providing individuals with the mechanisms needed to build trust into the IoT.
Yadav P, Moore J, Li Q, et al., 2018, Providing Occupancy as a Service with Databox, 1st ACM International Workshop on Smart Cities and Fog Computing (CitiFog), Publisher: ASSOC COMPUTING MACHINERY, Pages: 29-34
- Author Web Link
- Cite
- Citations: 5
Malekzadeh M, Clegg RG, Haddadi H, 2018, Replacement AutoEncoder: A Privacy-Preserving Algorithm for Sensory Data Analysis, 2018 IEEE/ACM THIRD INTERNATIONAL CONFERENCE ON INTERNET-OF-THINGS DESIGN AND IMPLEMENTATION (IOTDI 2020), Pages: 165-176
- Author Web Link
- Cite
- Citations: 22
Shamsabadi AS, Haddadi H, Cavallaro A, 2018, DISTRIBUTED ONE-CLASS LEARNING, Publisher: IEEE
- Author Web Link
- Open Access Link
- Cite
- Citations: 4
Katevas K, Tokarchuk L, Haddadi H, et al., 2017, Detecting Group Formations using iBeacon Technology, 15th ACM Annual International Conference on Mobile Systems, Applications, and Services (MobiSys), Publisher: ASSOC COMPUTING MACHINERY, Pages: 190-190
Hansel K, Haddadi H, Alomainy A, 2017, AWSense - A Framework for Collecting Sensing Data from the Apple Watch, 15th ACM Annual International Conference on Mobile Systems, Applications, and Services (MobiSys), Publisher: ASSOC COMPUTING MACHINERY, Pages: 188-188
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.