Search

Vection, the visual illusion of self-motion, provides a strong marker of the VR user experience and plays an important role in both presence and cybersickness. Traditional measurements have been conducted using questionnaires, which exhibit inherent limitations due to their subjective nature and preventing real-time adjustments. Detecting vection in real time would allow VR systems to adapt to users' needs, improving comfort and minimizing negative effects like motion sickness. This paper investigates the presence of vection markers in electroencephalogram (EEG) brain signals using evoked potentials (brain responses to external stimulations). We designed a VR experiment that induces vection using two conditions: (1) forward acceleration or (2) backward acceleration. We recorded both electroencephalographic (EEG) signals and gathered subjective reports on thirty (30) participants. We found an evoked potential of vection characterized by a positive peak around 600 ms (P600) after stimulus onset in the parietal region and a simultaneous negative peak in the frontal region. Our results also found participant variability in sensitivity to vection and cybersickness and EEG markers of acceleration across subjects. This result is promising for potential detection of vection using EEG and paves the way for future studies towards a better understanding of vection. It also provides insights into the functional role of the visual system and its integration with the vestibular system during motion-perception. It has the potential to help enhance VR user experience by qualifying users' perceived vection and adapting the VR environments accordingly.

Deep learning time-series models are often used to make forecasts that inform downstream decisions. Since these decisions can differ from those in the training set, there is an implicit requirement that time-series models will generalize outside of their training distribution. Despite this core requirement, time-series models are typically trained and evaluated on in-distribution predictive tasks. We extend the orthogonal statistical learning framework to train causal time-series models that generalize better when forecasting the effect of actions outside of their training distribution. To evaluate these models, we leverage Regression Discontinuity Designs popular in economics to construct a test set of causal treatment effects.

We propose Adaptive Randomized Smoothing (ARS) to certify the predictions of our test-time adaptive models against adversarial examples. ARS extends the analysis of randomized smoothing using $f$-Differential Privacy to certify the adaptive composition of multiple steps. For the first time, our theory covers the sound adaptive composition of general and high-dimensional functions of noisy inputs. We instantiate ARS on deep image classification to certify predictions against adversarial examples of bounded $L_{\infty}$ norm. In the $L_{\infty}$ threat model, ARS enables flexible adaptation through high-dimensional input-dependent masking. We design adaptivity benchmarks, based on CIFAR-10 and CelebA, and show that ARS improves standard test accuracy by $1$ to $15\%$ points. On ImageNet, ARS improves certified test accuracy by up to $1.6\%$ points over standard RS without adaptivity. Our code is available at https://github.com/ubc-systopia/adaptive-randomized-smoothing .

With the impending removal of third-party cookies from major browsers and the introduction of new privacy-preserving advertising APIs, the research community has a timely opportunity to assist industry in qualitatively improving the Web's privacy. This paper discusses our efforts, within a W3C community group, to enhance existing privacy-preserving advertising measurement APIs. We analyze designs from Google, Apple, Meta and Mozilla, and augment them with a more rigorous and efficient differential privacy (DP) budgeting component. Our approach, called Cookie Monster, enforces well-defined DP guarantees and enables advertisers to conduct more private measurement queries accurately. By framing the privacy guarantee in terms of an individual form of DP, we can make DP budgeting more efficient than in current systems that use a traditional DP definition. We incorporate Cookie Monster into Chrome and evaluate it on microbenchmarks and advertising datasets. Across workloads, Cookie Monster significantly outperforms baselines in enabling more advertising measurements under comparable DP protection., Comment: Published at SOSP '24. v5: typos and minor changes. v4: camera-ready version. v3: changed to non-anonymized name after acceptance notification, clarified text and reformatted graphs in {\S}8. v2: added pseudocode in {\S}3.3

We study Thompson Sampling-based algorithms for stochastic bandits with bounded rewards. As the existing problem-dependent regret bound for Thompson Sampling with Gaussian priors [Agrawal and Goyal, 2017] is vacuous when $T \le 288 e^{64}$, we derive a more practical bound that tightens the coefficient of the leading term %from $288 e^{64}$ to $1270$. Additionally, motivated by large-scale real-world applications that require scalability, adaptive computational resource allocation, and a balance in utility and computation, we propose two parameterized Thompson Sampling-based algorithms: Thompson Sampling with Model Aggregation (TS-MA-$\alpha$) and Thompson Sampling with Timestamp Duelling (TS-TD-$\alpha$), where $\alpha \in [0,1]$ controls the trade-off between utility and computation. Both algorithms achieve $O \left(K\ln^{\alpha+1}(T)/\Delta \right)$ regret bound, where $K$ is the number of arms, $T$ is the finite learning horizon, and $\Delta$ denotes the single round performance loss when pulling a sub-optimal arm.

We present PANORAMIA, a privacy leakage measurement framework for machine learning models that relies on membership inference attacks using generated data as non-members. By relying on generated non-member data, PANORAMIA eliminates the common dependency of privacy measurement tools on in-distribution non-member data. As a result, PANORAMIA does not modify the model, training data, or training process, and only requires access to a subset of the training data. We evaluate PANORAMIA on ML models for image and tabular data classification, as well as on large-scale language models., Comment: 36 pages

The Adam optimizer is a popular choice in contemporary deep learning, due to its strong empirical performance. However we observe that in privacy sensitive scenarios, the traditional use of Differential Privacy (DP) with the Adam optimizer leads to sub-optimal performance on several tasks. We find that this performance degradation is due to a DP bias in Adam's second moment estimator, introduced by the addition of independent noise in the gradient computation to enforce DP guarantees. This DP bias leads to a different scaling for low variance parameter updates, that is inconsistent with the behavior of non-private Adam. We propose DP-AdamBC, an optimization algorithm which removes the bias in the second moment estimation and retrieves the expected behaviour of Adam. Empirically, DP-AdamBC significantly improves the optimization performance of DP-Adam by up to 3.5% in final accuracy in image, text, and graph node classification tasks., Comment: Published as a conference paper at the 38th Annual AAAI Conference on Artificial Intelligence, Vancouver, 2024

The widespread adoption of encryption in network protocols has significantly improved the overall security of many Internet applications. However, these protocols cannot prevent network side-channel leaks -- leaks of sensitive information through the sizes and timing of network packets. We present NetShaper, a system that mitigates such leaks based on the principle of traffic shaping. NetShaper's traffic shaping provides differential privacy guarantees while adapting to the prevailing workload and congestion condition, and allows configuring a tradeoff between privacy guarantees, bandwidth and latency overheads. Furthermore, NetShaper provides a modular and portable tunnel endpoint design that can support diverse applications. We present a middlebox-based implementation of NetShaper and demonstrate its applicability in a video streaming and a web service application.

Differentially-private (DP) databases allow for privacy-preserving analytics over sensitive datasets or data streams. In these systems, user privacy is a limited resource that must be conserved with each query. We propose Turbo, a novel, state-of-the-art caching layer for linear query workloads over DP databases. Turbo builds upon private multiplicative weights (PMW), a DP mechanism that is powerful in theory but ineffective in practice, and transforms it into a highly-effective caching mechanism, PMW-Bypass, that uses prior query results obtained through an external DP mechanism to train a PMW to answer arbitrary future linear queries accurately and "for free" from a privacy perspective. Our experiments on public Covid19 and CitiBike datasets show that Turbo with PMW-Bypass conserves 1.7-15.9x more budget compared to vanilla PMW and simpler cache designs, a significant improvement. Moreover, Turbo provides support for range query workloads, such as timeseries or streams, where opportunities exist to further conserve privacy budget through DP parallel composition and warm-starting of PMW state. Our work provides a theoretical foundation and general system design for effective caching in DP databases., Comment: Extended version of a paper presented at the 29th ACM Symposium on Operating Systems Principles (SOSP '23)

We observe that the traditional use of DP with the Adam optimizer introduces a bias in the second moment estimation, due to the addition of independent noise in the gradient computation. This bias leads to a different scaling for low variance parameter updates, that is inconsistent with the behavior of non-private Adam, and Adam's sign descent interpretation. Empirically, correcting the bias introduced by DP noise significantly improves the optimization performance of DP-Adam., Comment: Published at ICLR 2023 Workshop on Trustworthy and Reliable Large-Scale Machine Learning Models

From small steps to great leaps, metaphors of spatial mobility abound to describe discovery processes. Here, we ground these ideas in formal terms by systematically studying scientific knowledge mobility patterns. We use low-dimensional embedding techniques to create a knowledge space made up of 1.5 million articles from the fields of physics, computer science, and mathematics. By analyzing the publication histories of individual researchers, we discover patterns of knowledge mobility that closely resemble physical mobility. In aggregate, the trajectories form mobility flows that can be described by a gravity model, with jumps more likely to occur in areas of high density and less likely to occur over longer distances. We identify two types of researchers from their individual mobility patterns: interdisciplinary explorers who pioneer new fields, and exploiters who are more likely to stay within their specific areas of expertise. Our results suggest that spatial mobility analysis is a valuable tool for understanding knowledge evolution., Comment: 15 pages, 5 figures, 10 Supplementary Figures

Machine learning (ML) models can leak information about users, and differential privacy (DP) provides a rigorous way to bound that leakage under a given budget. This DP budget can be regarded as a new type of compute resource in workloads of multiple ML models training on user data. Once it is used, the DP budget is forever consumed. Therefore, it is crucial to allocate it most efficiently to train as many models as possible. This paper presents the scheduler for privacy that optimizes for efficiency. We formulate privacy scheduling as a new type of multidimensional knapsack problem, called privacy knapsack, which maximizes DP budget efficiency. We show that privacy knapsack is NP-hard, hence practical algorithms are necessarily approximate. We develop an approximation algorithm for privacy knapsack, DPack, and evaluate it on microbenchmarks and on a new, synthetic private-ML workload we developed from the Alibaba ML cluster trace. We show that DPack: (1) often approaches the efficiency-optimal schedule, (2) consistently schedules more tasks compared to a state-of-the-art privacy scheduling algorithm that focused on fairness (1.3-1.7x in Alibaba, 1.0-2.6x in microbenchmarks), but (3) sacrifices some level of fairness for efficiency. Therefore, using DPack, DP ML operators should be able to train more models on the same amount of user data while offering the same privacy guarantee to their users., Comment: Published at EuroSys '25. v2: camera-ready version

Federated learning (FL) is an effective technique to directly involve edge devices in machine learning training while preserving client privacy. However, the substantial communication overhead of FL makes training challenging when edge devices have limited network bandwidth. Existing work to optimize FL bandwidth overlooks downstream transmission and does not account for FL client sampling. In this paper we propose GlueFL, a framework that incorporates new client sampling and model compression algorithms to mitigate low download bandwidths of FL clients. GlueFL prioritizes recently used clients and bounds the number of changed positions in compression masks in each round. Across three popular FL datasets and three state-of-the-art strategies, GlueFL reduces downstream client bandwidth by 27% on average and reduces training time by 29% on average.

Sparse decision trees are one of the most common forms of interpretable models. While recent advances have produced algorithms that fully optimize sparse decision trees for prediction, that work does not address policy design, because the algorithms cannot handle weighted data samples. Specifically, they rely on the discreteness of the loss function, which means that real-valued weights cannot be directly used. For example, none of the existing techniques produce policies that incorporate inverse propensity weighting on individual data points. We present three algorithms for efficient sparse weighted decision tree optimization. The first approach directly optimizes the weighted loss function; however, it tends to be computationally inefficient for large datasets. Our second approach, which scales more efficiently, transforms weights to integer values and uses data duplication to transform the weighted decision tree optimization problem into an unweighted (but larger) counterpart. Our third algorithm, which scales to much larger datasets, uses a randomized procedure that samples each data point with a probability proportional to its weight. We present theoretical bounds on the error of the two fast methods and show experimentally that these methods can be two orders of magnitude faster than the direct optimization of the weighted loss, without losing significant accuracy., Comment: Advances in Interpretable Machine Learning, AIMLAI 2022. arXiv admin note: text overlap with arXiv:2112.00798

Pseudo-haptic techniques are used to modify haptic perception by appropriately changing visual feedback to body movements. Based on the knowledge that tendon vibration can affect our somatosensory perception, this paper proposes a method for leveraging tendon vibration to enhance pseudo-haptics during free arm motion. Three experiments were performed to examine the impact of tendon vibration on the range and resolution of pseudo-haptics. The first experiment investigated the effect of tendon vibration on the detection threshold of the discrepancy between visual and physical motion. The results indicated that vibrations applied to the inner tendons of the wrist and elbow increased the threshold, suggesting that tendon vibration can augment the applicable visual motion gain by approximately 13\% without users detecting the visual/physical discrepancy. Furthermore, the results demonstrate that tendon vibration acts as noise on haptic motion cues. The second experiment assessed the impact of tendon vibration on the resolution of pseudo-haptics by determining the just noticeable difference in pseudo-weight perception. The results suggested that the tendon vibration does not largely compromise the resolution of pseudo-haptics. The third experiment evaluated the equivalence between the weight perception triggered by tendon vibration and that by visual motion gain, that is, the point of subjective equality. The results revealed that vibration amplifies the weight perception and its effect was equivalent to that obtained using a gain of 0.64 without vibration, implying that the tendon vibration also functions as an additional haptic cue. Our results provide design guidelines and future work for enhancing pseudo-haptics with tendon vibration., Comment: This paper has been accepted by IEEE TVCG

We develop a new, principled algorithm for estimating the contribution of training data points to the behavior of a deep learning model, such as a specific prediction it makes. Our algorithm estimates the AME, a quantity that measures the expected (average) marginal effect of adding a data point to a subset of the training data, sampled from a given distribution. When subsets are sampled from the uniform distribution, the AME reduces to the well-known Shapley value. Our approach is inspired by causal inference and randomized experiments: we sample different subsets of the training data to train multiple submodels, and evaluate each submodel's behavior. We then use a LASSO regression to jointly estimate the AME of each data point, based on the subset compositions. Under sparsity assumptions ($k \ll N$ datapoints have large AME), our estimator requires only $O(k\log N)$ randomized submodel trainings, improving upon the best prior Shapley value estimators., Comment: ICML 2022

Walking-in-place (WIP) is a locomotion technique that enables users to "walk infinitely" through vast virtual environments using walking-like gestures within a limited physical space. This paper investigates alternative interaction schemes for WIP, addressing successively the control, input, and output of WIP. First, we introduce a novel height-based control to increase advanced speed. Second, we introduce a novel input system for WIP based on elastic and passive strips. Third, we introduce the use of pseudo-haptic feedback as a novel output for WIP meant to alter walking sensations. The results of a series of user studies show that height and frequency based control of WIP can facilitate higher virtual speed with greater efficacy and ease than in frequency-based WIP. Second, using an upward elastic input system can result in a stable virtual speed control, although excessively strong elastic forces may impact the usability and user experience. Finally, using a pseudo-haptic approach can improve the perceived realism of virtual slopes. Taken together, our results suggest that, for future VR applications, there is value in further research into the use of alternative interaction schemes for walking-in-place., Comment: This paper has been accepted by IEEE TVCG

Listing triangles is a fundamental graph problem with many applications, and large graphs require fast algorithms. Vertex ordering allows the orientation of edges from lower to higher vertex indices, and state-of-the-art triangle listing algorithms use this to accelerate their execution and to bound their time complexity. Yet, only basic orderings have been tested. In this paper, we show that studying the precise cost of algorithms instead of their bounded complexity leads to faster solutions. We introduce cost functions that link ordering properties with the running time of a given algorithm. We prove that their minimization is NP-hard and propose heuristics to obtain new orderings with different trade-offs between cost reduction and ordering time. Using datasets with up to two billion edges, we show that our heuristics accelerate the listing of triangles by an average of 38% when the ordering is already given as an input, and 16% when the ordering time is included., Comment: 11 pages, 4 figures. Open-source C++ code available at: https://github.com/lecfab/volt

We observe that many system policies that make threshold decisions involving a resource (e.g., time, memory, cores) naturally reveal additional, or implicit feedback. For example, if a system waits X min for an event to occur, then it automatically learns what would have happened if it waited

Autonomous vision-based spaceborne navigation is an enabling technology for future on-orbit servicing and space logistics missions. While computer vision in general has benefited from Machine Learning (ML), training and validating spaceborne ML models are extremely challenging due to the impracticality of acquiring a large-scale labeled dataset of images of the intended target in the space environment. Existing datasets, such as Spacecraft PosE Estimation Dataset (SPEED), have so far mostly relied on synthetic images for both training and validation, which are easy to mass-produce but fail to resemble the visual features and illumination variability inherent to the target spaceborne images. In order to bridge the gap between the current practices and the intended applications in future space missions, this paper introduces SPEED+: the next generation spacecraft pose estimation dataset with specific emphasis on domain gap. In addition to 60,000 synthetic images for training, SPEED+ includes 9,531 hardware-in-the-loop images of a spacecraft mockup model captured from the Testbed for Rendezvous and Optical Navigation (TRON) facility. TRON is a first-of-a-kind robotic testbed capable of capturing an arbitrary number of target images with accurate and maximally diverse pose labels and high-fidelity spaceborne illumination conditions. SPEED+ is used in the second international Satellite Pose Estimation Challenge co-hosted by SLAB and the Advanced Concepts Team of the European Space Agency to evaluate and compare the robustness of spaceborne ML models trained on synthetic images.

The sensation of self-motion is essential in many virtual reality applications, from entertainment to training, such as flying and driving simulators. If the common approach used in amusement parks is to actuate the seats with cumbersome systems, multisensory integration can also be leveraged to get rich effects from lightweight solutions. In this short paper, we introduce a novel approach called the "Kinesthetic HMD": actuating a head-mounted display with force feedback in order to provide sensations of self-motion. We discuss its design considerations and demonstrate an augmented flight simulator use case with a proof-of-concept prototype. We conducted a user study assessing our approach's ability to enhance self-motion sensations. Taken together, our results show that our Kinesthetic HMD provides significantly stronger and more egocentric sensations than a visual-only self-motion experience. Thus, by providing congruent vestibular and proprioceptive cues related to balance and self-motion, the Kinesthetic HMD represents a promising approach for a variety of virtual reality applications in which motion sensations are prominent., Comment: 8 pages, 7 figures

Machine learning (ML) models trained on personal data have been shown to leak information about users. Differential privacy (DP) enables model training with a guaranteed bound on this leakage. Each new model trained with DP increases the bound on data leakage and can be seen as consuming part of a global privacy budget that should not be exceeded. This budget is a scarce resource that must be carefully managed to maximize the number of successfully trained models. We describe PrivateKube, an extension to the popular Kubernetes datacenter orchestrator that adds privacy as a new type of resource to be managed alongside other traditional compute resources, such as CPU, GPU, and memory. The abstractions we design for the privacy resource mirror those defined by Kubernetes for traditional resources, but there are also major differences. For example, traditional compute resources are replenishable while privacy is not: a CPU can be regained after a model finishes execution while privacy budget cannot. This distinction forces a re-design of the scheduler. We present DPF (Dominant Private Block Fairness) -- a variant of the popular Dominant Resource Fairness (DRF) algorithm -- that is geared toward the non-replenishable privacy resource but enjoys similar theoretical properties as DRF. We evaluate PrivateKube and DPF on microbenchmarks and an ML workload on Amazon Reviews data. Compared to existing baselines, DPF allows training more models under the same global privacy guarantee. This is especially true for DPF over R\'enyi DP, a highly composable form of DP., Comment: Extended version of a paper presented at the 15th USENIX Symposium on Operating Systems Design and Implementation (OSDI '21)