1 code implementation • 20 Oct 2023 • Mohammad Malekzadeh, Fahim Kawsar
In split inference, a deep neural network (DNN) is partitioned to run the early part of the DNN at the edge and the later part of the DNN in the cloud.
1 code implementation • 31 Jul 2023 • Shohreh Deldari, Dimitris Spathis, Mohammad Malekzadeh, Fahim Kawsar, Flora Salim, Akhil Mathur
Limited availability of labeled data for machine learning on multimodal time-series extensively hampers progress in the field.
1 code implementation • 8 Dec 2022 • Mohammad Malekzadeh, Deniz Gunduz
Privacy-preserving inference in edge computing paradigms encourages the users of machine-learning services to locally run a model on their private input, for a target task, and only share the model's outputs with the server.
1 code implementation • 8 Nov 2022 • Fan Mo, Mohammad Malekzadeh, Soumyajit Chatterjee, Fahim Kawsar, Akhil Mathur
Federated learning (FL) in multidevice environments creates new opportunities to learn from a vast and diverse amount of private data.
1 code implementation • 9 Aug 2021 • Aman Priyanshu, Rakshit Naidu, FatemehSadat Mireshghallah, Mohammad Malekzadeh
Tuning the hyperparameters in the differentially private stochastic gradient descent (DPSGD) is a fundamental challenge.
no code implementations • 28 May 2021 • Fan Mo, Anastasia Borovykh, Mohammad Malekzadeh, Soteris Demetriou, Deniz Gündüz, Hamed Haddadi
Our proposed framework enables clients to localize and quantify the private information leakage in a layer-wise manner, and enables a better understanding of the sources of information leakage in collaborative learning, which can be used by future studies to benchmark new attacks and defense mechanisms.
1 code implementation • 25 May 2021 • Mohammad Malekzadeh, Anastasia Borovykh, Deniz Gündüz
It is known that deep neural networks, trained for the classification of non-sensitive target attributes, can reveal sensitive attributes of their input data through internal representations extracted by the classifier.
1 code implementation • 27 Jan 2021 • Mohammad Malekzadeh, Burak Hasircioglu, Nitish Mital, Kunal Katarya, Mehmet Emre Ozfatura, Deniz Gündüz
While rich medical datasets are hosted in hospitals distributed across the world, concerns on patients' privacy is a barrier against using such data to train deep neural networks (DNNs) for medical diagnostics.
no code implementations • 17 Oct 2020 • Fan Mo, Anastasia Borovykh, Mohammad Malekzadeh, Hamed Haddadi, Soteris Demetriou
Training deep neural networks via federated learning allows clients to share, instead of the original data, only the model trained on their data.
no code implementations • 4 Sep 2020 • Giuseppe Siracusano, Salvator Galea, Davide Sanvito, Mohammad Malekzadeh, Hamed Haddadi, Gianni Antichi, Roberto Bifulco
In this paper we show that the data plane of commodity programmable (Network Interface Cards) NICs can run neural network inference tasks required by packet monitoring applications, with low overhead.
2 code implementations • 5 Aug 2020 • Mohammad Malekzadeh, Richard G. Clegg, Andrea Cavallaro, Hamed Haddadi
We introduce a dimension-adaptive pooling (DAP) layer that makes DNNs flexible and more robust to changes in sensor availability and in sampling rate.
1 code implementation • 14 Nov 2019 • Mohammad Malekzadeh, Richard G. Clegg, Andrea Cavallaro, Hamed Haddadi
Sensitive inferences and user re-identification are major threats to privacy when raw sensor data from wearable or portable devices are shared with cloud-assisted applications.
1 code implementation • 10 Sep 2019 • Mohammad Malekzadeh, Dimitrios Athanasakis, Hamed Haddadi, Benjamin Livshits
Contextual bandit algorithms~(CBAs) often rely on personal data to provide recommendations.
1 code implementation • 21 Jun 2019 • Edoardo Lisi, Mohammad Malekzadeh, Hamed Haddadi, F. Din-Houn Lau, Seth Flaxman
Realisations from this distribution can be used by the CGAN to generate "future" paintings.
1 code implementation • 26 Oct 2018 • Mohammad Malekzadeh, Richard G. Clegg, Andrea Cavallaro, Hamed Haddadi
Motion sensors such as accelerometers and gyroscopes measure the instant acceleration and rotation of a device, in three dimensions.
1 code implementation • 21 Feb 2018 • Mohammad Malekzadeh, Richard G. Clegg, Andrea Cavallaro, Hamed Haddadi
Results show that the proposed framework maintains the usefulness of the transformed data for activity recognition, with an average loss of only around three percentage points, while reducing the possibility of gender classification to around 50\%, the target random guess, from more than 90\% when using raw sensor data.
1 code implementation • 18 Oct 2017 • Mohammad Malekzadeh, Richard G. Clegg, Hamed Haddadi
Though access to the sensory data is critical to the success of many beneficial applications such as health monitoring or activity recognition, a wide range of potentially sensitive information about the individuals can also be discovered through access to sensory data and this cannot easily be protected using traditional privacy approaches.