Yushan Huang, Josh Millar, Yuxuan Long, Yuchen Zhao, Hamed Haddadi
The personalization of machine learning (ML) models to address data drift is a significant challenge in the context of Internet of Things (IoT) applications. Presently, most approaches focus on fine-tuning either the full base model or its last few layers to adapt to new data, while often neglecting energy costs. However, various types of data drift exist, and fine-tuning the full base model or the last few layers may not result in optimal performance in certain scenarios. We propose Target Block Fine-Tuning (TBFT), a low-energy adaptive personalization framework designed for resource-constrained devices. We categorize data drift and personalization into three types: input-level, feature-level, and output-level. For each type, we fine-tune different blocks of the model to achieve optimal performance with reduced energy costs. Specifically, input-, feature-, and output-level correspond to fine-tuning the front, middle, and rear blocks of the model. We evaluate TBFT on a ResNet model, three datasets, three different training sizes, and a Raspberry Pi. Compared with the $Block Avg$, where each block is fine-tuned individually and their performance improvements are averaged, TBFT exhibits an improvement in model accuracy by an average of 15.30% whilst saving 41.57% energy consumption on average compared with full fine-tuning.
Sandra Siby, Sina Abdollahi, Mohammad Maheri, Marios Kogias, Hamed Haddadi
Machine-learning (ML) models are increasingly being deployed on edge devices to provide a variety of services. However, their deployment is accompanied by challenges in model privacy and auditability. Model providers want to ensure that (i) their proprietary models are not exposed to third parties; and (ii) be able to get attestations that their genuine models are operating on edge devices in accordance with the service agreement with the user. Existing measures to address these challenges have been hindered by issues such as high overheads and limited capability (processing/secure memory) on edge devices. In this work, we propose GuaranTEE, a framework to provide attestable private machine learning on the edge. GuaranTEE uses Confidential Computing Architecture (CCA), Arm's latest architectural extension that allows for the creation and deployment of dynamic Trusted Execution Environments (TEEs) within which models can be executed. We evaluate CCA's feasibility to deploy ML models by developing, evaluating, and openly releasing a prototype. We also suggest improvements to CCA to facilitate its use in protecting the entire ML deployment pipeline on edge devices.
Salim Chouaki, Oana Goga, Hamed Haddadi, Peter Snyder
We present the first extensive measurement of the privacy properties of the advertising systems used by privacy-focused search engines. We propose an automated methodology to study the impact of clicking on search ads on three popular private search engines which have advertising-based business models: StartPage, Qwant, and DuckDuckGo, and we compare them to two dominant data-harvesting ones: Google and Bing. We investigate the possibility of third parties tracking users when clicking on ads by analyzing first-party storage, redirection domain paths, and requests sent before, when, and after the clicks. Our results show that privacy-focused search engines fail to protect users' privacy when clicking ads. Users' requests are sent through redirectors on 4% of ad clicks on Bing, 86% of ad clicks on Qwant, and 100% of ad clicks on Google, DuckDuckGo, and StartPage. Even worse, advertising systems collude with advertisers across all search engines by passing unique IDs to advertisers in most ad clicks. These IDs allow redirectors to aggregate users' activity on ads' destination websites in addition to the activity they record when users are redirected through them. Overall, we observe that both privacy-focused and traditional search engines engage in privacy-harming behaviors allowing cross-site tracking, even in privacy-enhanced browsers.
Yuting Zhan, Hamed Haddadi, Afra Mashhadi
As mobile devices and location-based services are increasingly developed in different smart city scenarios and applications, many unexpected privacy leakages have arisen due to geolocated data collection and sharing. User re-identification and other sensitive inferences are major privacy threats when geolocated data are shared with cloud-assisted applications. Significantly, four spatio-temporal points are enough to uniquely identify 95\% of the individuals, which exacerbates personal information leakages. To tackle malicious purposes such as user re-identification, we propose an LSTM-based adversarial mechanism with representation learning to attain a privacy-preserving feature representation of the original geolocated data (i.e., mobility data) for a sharing purpose. These representations aim to maximally reduce the chance of user re-identification and full data reconstruction with a minimal utility budget (i.e., loss). We train the mechanism by quantifying privacy-utility trade-off of mobility datasets in terms of trajectory reconstruction risk, user re-identification risk, and mobility predictability. We report an exploratory analysis that enables the user to assess this trade-off with a specific loss function and its weight parameters. The extensive comparison results on four representative mobility datasets demonstrate the superiority of our proposed architecture in mobility privacy protection and the efficiency of the proposed privacy-preserving features extractor. We show that the privacy of mobility traces attains decent protection at the cost of marginal mobility utility. Our results also show that by exploring the Pareto optimal setting, we can simultaneously increase both privacy (45%) and utility (32%).
Savvas Hadjixenophontos, Anna Maria Mandalari, Yuchen Zhao, Hamed Haddadi
Consumer healthcare Internet of Things (IoT) devices are gaining popularity in our homes and hospitals. These devices provide continuous monitoring at a low cost and can be used to augment high-precision medical equipment. However, major challenges remain in applying pre-trained global models for anomaly detection on smart health monitoring, for a diverse set of individuals that they provide care for. In this paper, we propose PRISM, an edge-based system for experimenting with in-home smart healthcare devices. We develop a rigorous methodology that relies on automated IoT experimentation. We use a rich real-world dataset from in-home patient monitoring from 44 households of People Living With Dementia (PLWD) over two years. Our results indicate that anomalies can be identified with accuracy up to 99% and mean training times as low as 0.88 seconds. While all models achieve high accuracy when trained on the same patient, their accuracy degrades when evaluated on different patients.
Yuting Zhan, Hamed Haddadi, Afra Mashhadi
Preserving the individuals' privacy in sharing spatial-temporal datasets is critical to prevent re-identification attacks based on unique trajectories. Existing privacy techniques tend to propose ideal privacy-utility tradeoffs, however, largely ignore the fairness implications of mobility models and whether such techniques perform equally for different groups of users. The quantification between fairness and privacy-aware models is still unclear and there barely exists any defined sets of metrics for measuring fairness in the spatial-temporal context. In this work, we define a set of fairness metrics designed explicitly for human mobility, based on structural similarity and entropy of the trajectories. Under these definitions, we examine the fairness of two state-of-the-art privacy-preserving models that rely on GAN and representation learning to reduce the re-identification rate of users for data sharing. Our results show that while both models guarantee group fairness in terms of demographic parity, they violate individual fairness criteria, indicating that users with highly similar trajectories receive disparate privacy gain. We conclude that the tension between the re-identification task and individual fairness needs to be considered for future spatial-temporal data analysis and modelling to achieve a privacy-preserving fairness-aware setting.
Edoardo Lisi, Mohammad Malekzadeh, Hamed Haddadi, F. Din-Houn Lau, Seth Flaxman
Conditional Generative Adversarial Networks~(CGAN) are a recent and popular method for generating samples from a probability distribution conditioned on latent information. The latent information often comes in the form of a discrete label from a small set. We propose a novel method for training CGANs which allows us to condition on a sequence of continuous latent distributions $f^{(1)}, \ldots, f^{(K)}$. This training allows CGANs to generate samples from a sequence of distributions. We apply our method to paintings from a sequence of artistic movements, where each movement is considered to be its own distribution. Exploiting the temporal aspect of the data, a vector autoregressive (VAR) model is fitted to the means of the latent distributions that we learn, and used for one-step-ahead forecasting, to predict the latent distribution of a future art movement $f^{(K+1)}$. Realisations from this distribution can be used by the CGAN to generate "future" paintings. In experiments, this novel methodology generates accurate predictions of the evolution of art. The training set consists of a large dataset of past paintings. While there is no agreement on exactly what current art period we find ourselves in, we test on plausible candidate sets of present art, and show that the mean distance to our predictions is small.
Michael Smith, Peter Snyder, Moritz Haller, Benjamin Livshits, Deian Stefan, Hamed Haddadi
A core problem in the development and maintenance of crowd-sourced filter lists is that their maintainers cannot confidently predict whether (and where) a new filter list rule will break websites. This is a result of enormity of the Web, which prevents filter list authors from broadly understanding the impact of a new blocking rule before they ship it to millions of users. The inability of filter list authors to evaluate the Web compatibility impact of a new rule before shipping it severely reduces the benefits of filter-list-based content blocking: filter lists are both overly-conservative (i.e. rules are tailored narrowly to reduce the risk of breaking things) and error-prone (i.e. blocking tools still break large numbers of sites). To scale to the size and scope of the Web, filter list authors need an automated system to detect when a new filter rule breaks websites, before that breakage has a chance to make it to end users. In this work, we design and implement the first automated system for predicting when a filter list rule breaks a website. We build a classifier, trained on a dataset generated by a combination of compatibility data from the EasyList project and novel browser instrumentation, and find it is accurate to practical levels (AUC 0.88). Our open source system requires no human interaction when assessing the compatibility risk of a proposed privacy intervention. We also present the 40 page behaviors that most predict breakage in observed websites.
Yuchen Zhao, Sayed Saad Afzal, Waleed Akbar, Osvy Rodriguez, Fan Mo, David Boyle, Fadel Adib, Hamed Haddadi
This paper is motivated by a simple question: Can we design and build battery-free devices capable of machine learning and inference in underwater environments? An affirmative answer to this question would have significant implications for a new generation of underwater sensing and monitoring applications for environmental monitoring, scientific exploration, and climate/weather prediction. To answer this question, we explore the feasibility of bridging advances from the past decade in two fields: battery-free networking and low-power machine learning. Our exploration demonstrates that it is indeed possible to enable battery-free inference in underwater environments. We designed a device that can harvest energy from underwater sound, power up an ultra-low-power microcontroller and on-board sensor, perform local inference on sensed measurements using a lightweight Deep Neural Network, and communicate the inference result via backscatter to a receiver. We tested our prototype in an emulated marine bioacoustics application, demonstrating the potential to recognize underwater animal sounds without batteries. Through this exploration, we highlight the challenges and opportunities for making underwater battery-free inference and machine learning ubiquitous.
Matteo Varvello, Kleomenis Katevas, Mihai Plesa, Hamed Haddadi, Fabian Bustamante, Ben Livshits
Advances in cloud computing have simplified the way that both software development and testing are performed. This is not true for battery testing for which state of the art test-beds simply consist of one phone attached to a power meter. These test-beds have limited resources, access, and are overall hard to maintain; for these reasons, they often sit idle with no experiment to run. In this paper, we propose to share existing battery testbeds and transform them into vantage points of BatteryLab, a power monitoring platform offering heterogeneous devices and testing conditions. We have achieved this vision with a combination of hardware and software which allow to augment existing battery test-beds with remote capabilities. BatteryLab currently counts three vantage points, one in Europe and two in the US, hosting three Android devices and one iPhone 7. We benchmark BatteryLab with respect to the accuracy of its battery readings, system performance, and platform heterogeneity. Next, we demonstrate how measurements can be run atop of BatteryLab by developing the "Web Power Monitor" (WPM), a tool which can measure website power consumption at scale. We released WPM and used it to report on the energy consumption of Alexa's top 1,000 websites across 3 locations and 4 devices (both Android and iOS).
Yuting Zhan, Hamed Haddadi
Smart home solutions increasingly rely on a variety of sensors for behavioral analytics and activity recognition to provide context-aware applications and personalized care. Optimizing the sensor network is one of the most important approaches to ensure classification accuracy and the system's efficiency. However, the trade-off between the cost and performance is often a challenge in real deployments, particularly for multiple-occupancy smart homes or care homes. In this paper, using real indoor activity and mobility traces, floor plans, and synthetic multi-occupancy behavior models, we evaluate several multi-occupancy household scenarios with 2-5 residents. We explore and quantify the trade-offs between the cost of sensor deployments and expected labeling accuracy in different scenarios. Our evaluation across different scenarios show that the performance of the desired context-aware task is affected by different localization resolutions, the number of residents, the number of sensors, and varying sensor deployments. To aid in accelerating the adoption of practical sensor-based activity recognition technology, we design MoSen, a framework to simulate the interaction dynamics between sensor-based environments and multiple residents. By evaluating the factors that affect the performance of the desired sensor network, we provide a sensor selection strategy and design metrics for sensor layout in real environments. Using our selection strategy in a 5-person scenario case study, we demonstrate that MoSen can significantly improve overall system performance without increasing the deployment costs.
Lorenzo Minto, Moritz Haller, Hamed Haddadi, Benjamin Livshits
Recommender systems are commonly trained on centrally collected user interaction data like views or clicks. This practice however raises serious privacy concerns regarding the recommender's collection and handling of potentially sensitive data. Several privacy-aware recommender systems have been proposed in recent literature, but comparatively little attention has been given to systems at the intersection of implicit feedback and privacy. To address this shortcoming, we propose a practical federated recommender system for implicit data under user-level local differential privacy (LDP). The privacy-utility trade-off is controlled by parameters $ε$ and $k$, regulating the per-update privacy budget and the number of $ε$-LDP gradient updates sent by each user respectively. To further protect the user's privacy, we introduce a proxy network to reduce the fingerprinting surface by anonymizing and shuffling the reports before forwarding them to the recommender. We empirically demonstrate the effectiveness of our framework on the MovieLens dataset, achieving up to Hit Ratio with K=10 (HR@10) 0.68 on 50k users with 5k items. Even on the full dataset, we show that it is possible to achieve reasonable utility with HR@10>0.5 without compromising user privacy.
Chaoyun Zhang, Paul Patras, Hamed Haddadi
The rapid uptake of mobile devices and the rising popularity of mobile applications and services pose unprecedented demands on mobile and wireless networking infrastructure. Upcoming 5G systems are evolving to support exploding mobile traffic volumes, agile management of network resource to maximize user experience, and extraction of fine-grained real-time analytics. Fulfilling these tasks is challenging, as mobile environments are increasingly complex, heterogeneous, and evolving. One potential solution is to resort to advanced machine learning techniques to help managing the rise in data volumes and algorithm-driven applications. The recent success of deep learning underpins new and powerful tools that tackle problems in this space. In this paper we bridge the gap between deep learning and mobile and wireless networking research, by presenting a comprehensive survey of the crossovers between the two areas. We first briefly introduce essential background and state-of-the-art in deep learning techniques with potential applications to networking. We then discuss several techniques and platforms that facilitate the efficient deployment of deep learning onto mobile systems. Subsequently, we provide an encyclopedic review of mobile and wireless networking research based on deep learning, which we categorize by different domains. Drawing from our experience, we discuss how to tailor deep learning to mobile environments. We complete this survey by pinpointing current challenges and open future directions for research.
Ali Shahin Shamsabadi, Hamed Haddadi, Andrea Cavallaro
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 autoencoders. 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.
Seyed Ali Osia, Ali Shahin Shamsabadi, Sina Sajadmanesh, Ali Taheri, Kleomenis Katevas, Hamid R. Rabiee, Nicholas D. Lane, Hamed Haddadi
Internet of Things (IoT) devices and applications are being deployed in our homes and workplaces. These devices often rely on continuous data collection to feed machine learning models. However, this approach introduces several privacy and efficiency challenges, as the service operator can perform unwanted inferences on the available data. Recently, advances in edge processing have paved the way for more efficient, and private, data processing at the source for simple tasks and lighter models, though they remain a challenge for larger, and more complicated models. In this paper, we present a hybrid approach for breaking down large, complex deep neural networks for cooperative, privacy-preserving analytics. To this end, instead of performing the whole operation on the cloud, we let an IoT device to run the initial layers of the neural network, and then send the output to the cloud to feed the remaining layers and produce the final result. In order to ensure that the user's device contains no extra information except what is necessary for the main task and preventing any secondary inference on the data, we introduce Siamese fine-tuning. We evaluate the privacy benefits of this approach based on the information exposed to the cloud service. We also assess the local inference cost of different layers on a modern handset. Our evaluations show that by using Siamese fine-tuning and at a small processing cost, we can greatly reduce the level of unnecessary, potentially sensitive information in the personal data, and thus achieving the desired trade-off between utility, privacy, and performance.
Mir Masood Ali, Peter Snyder, Chris Kanich, Hamed Haddadi
This work presents Unbundle-Rewrite-Rebundle (URR), a system for detecting privacy-harming portions of bundled JavaScript code and rewriting that code at runtime to remove the privacy-harming behavior without breaking the surrounding code or overall application. URR is a novel solution to the problem of JavaScript bundles, where websites pre-compile multiple code units into a single file, making it impossible for content filters and ad-blockers to differentiate between desired and unwanted resources. Where traditional content filtering tools rely on URLs, URR analyzes the code at the AST level, and replaces harmful AST sub-trees with privacy-and-functionality maintaining alternatives. We present an open-sourced implementation of URR as a Firefox extension and evaluate it against JavaScript bundles generated by the most popular bundling system (Webpack) deployed on the Tranco 10k. We evaluate URR by precision (1.00), recall (0.95), and speed (0.43s per script) when detecting and rewriting three representative privacy-harming libraries often included in JavaScript bundles, and find URR to be an effective approach to a large-and-growing blind spot unaddressed by current privacy tools.
Hamed Haddadi, Raul Landa, Miguel Rio, Saleem Bhatti
The high volume of packets and packet rates of traffic on some router links makes it exceedingly difficult for routers to examine every packet in order to keep detailed statistics about the traffic which is traversing the router. Sampling is commonly applied on routers in order to limit the load incurred by the collection of information that the router has to undertake when evaluating flow information for monitoring purposes. The sampling process in nearly all cases is a deterministic process of choosing 1 in every N packets on a per-interface basis, and then forming the flow statistics based on the collected sampled statistics. Even though this sampling may not be significant for some statistics, such as packet rate, others can be severely distorted. However, it is important to consider the sampling techniques and their relative accuracy when applied to different traffic patterns. The main disadvantage of sampling is the loss of accuracy in the collected trace when compared to the original traffic stream. To date there has not been a detailed analysis of the impact of sampling at a router in various traffic profiles and flow criteria. In this paper, we assess the performance of the sampling process as used in NetFlow in detail, and we discuss some techniques for the compensation of loss of monitoring detail.
Mohammad M Maheri, Hamed Haddadi, Alex Davidson
Verification of the integrity of deep learning inference is crucial for understanding whether a model is being applied correctly. However, such verification typically requires access to model weights and (potentially sensitive or private) training data. So-called Zero-knowledge Succinct Non-Interactive Arguments of Knowledge (ZK-SNARKs) would appear to provide the capability to verify model inference without access to such sensitive data. However, applying ZK-SNARKs to modern neural networks, such as transformers and large vision models, introduces significant computational overhead. We present TeleSparse, a ZK-friendly post-processing mechanisms to produce practical solutions to this problem. TeleSparse tackles two fundamental challenges inherent in applying ZK-SNARKs to modern neural networks: (1) Reducing circuit constraints: Over-parameterized models result in numerous constraints for ZK-SNARK verification, driving up memory and proof generation costs. We address this by applying sparsification to neural network models, enhancing proof efficiency without compromising accuracy or security. (2) Minimizing the size of lookup tables required for non-linear functions, by optimizing activation ranges through neural teleportation, a novel adaptation for narrowing activation functions' range. TeleSparse reduces prover memory usage by 67% and proof generation time by 46% on the same model, with an accuracy trade-off of approximately 1%. We implement our framework using the Halo2 proving system and demonstrate its effectiveness across multiple architectures (Vision-transformer, ResNet, MobileNet) and datasets (ImageNet,CIFAR-10,CIFAR-100). This work opens new directions for ZK-friendly model design, moving toward scalable, resource-efficient verifiable deep learning.
Alisha Ukani, Hamed Haddadi, Ali Shahin Shamsabadi, Peter Snyder
This paper presents a systematic evaluation of the privacy behaviors and attributes of eight recent, popular browser agents. Browser agents are software that automate Web browsing using large language models and ancillary tooling. However, the automated capabilities that make browser agents powerful also make them high-risk points of failure. Both the kinds of tasks browser agents are designed to execute, along with the kinds of information browser agents are entrusted with to fulfill those tasks, mean that vulnerabilities in these tools can result in enormous privacy harm. This work presents a framework of five broad factors (totaling 15 distinct measurements) to measure the privacy risks in browser agents. Our framework assesses i. vulnerabilities in the browser agent's components, ii. how the browser agent protects against website behaviors, iii. whether the browser agent prevents cross-site tracking, iv. how the agent responds to privacy-affecting prompts, and v. whether the tool leaks personal information to sites. We apply our framework to eight browser agents and identify 30 vulnerabilities, ranging from disabled browser privacy features to "autocompleting" sensitive personal information in form fields. We have responsibly disclosed our findings, and plan to release our dataset and other artifacts.
Javad Forough, Hamed Haddadi
Membership Inference Attacks (MIAs) expose privacy risks by determining whether a specific sample was part of a model's training set. These threats are especially serious in sensitive domains such as healthcare and finance. Traditional mitigation techniques, such as static differential privacy, rely on injecting a fixed amount of noise during training or inference. However, this often leads to a detrimental trade-off: the noise may be insufficient to counter sophisticated attacks or, when increased, can substantially degrade model accuracy. To address this limitation, we propose DynaNoise, an adaptive inference-time defense that modulates injected noise based on per-query sensitivity. DynaNoise estimates risk using measures such as Shannon entropy and scales the noise variance accordingly, followed by a smoothing step that re-normalizes the perturbed outputs to preserve predictive utility. We further introduce MIDPUT (Membership Inference Defense Privacy-Utility Trade-off), a scalar metric that captures both privacy gains and accuracy retention. Our evaluation on several benchmark datasets demonstrates that DynaNoise substantially lowers attack success rates while maintaining competitive accuracy, achieving strong overall MIDPUT scores compared to state-of-the-art defenses.