Hugo Latapie

Crowd monitoring in sports stadiums is essential for public safety and audience experience. Traditional methods rely on manual observation, cameras, and microphones, which can be disruptive and raise privacy concerns. Floor vibration sensing offers a non-intrusive alternative, but the lack of open-source vibration datasets limits data-driven model development. In this paper, we propose Vi bration L everages A udio (ViLA) , a vibration-based crowd monitoring framework that reduces reliance on labeled vibration data through cross-modal pre-training. Specifically, ViLA is first pre-trained on general-purpose audio data in an unsupervised manner, then fine-tuned using a small amount of labeled vibration data. This approach enables the model to learn general spectral representations from audio and transfer this knowledge to the vibration domain, which is especially useful in environments with high variability and limited sensing duration, such as live sports events. We evaluate ViLA in two stadium environments across different sports. We also analyze audio–vibration relationships to understand how pre-training influences model performance. In a preliminary single-run modality comparison, general-purpose audio produced the highest accuracy among the evaluated pretraining modalities. Across five-seed same-day evaluations, audio pretraining improved performance in some games, particularly when the scratch baseline was weak, while the magnitude of benefit varied across games.

Federated learning (FL) is a distributed learning paradigm that facilitates training a global machine-learning model without collecting the raw data from distributed clients. Recent advances in FL have addressed several considerations that are likely to transpire in realistic settings, such as data distribution heterogeneity among clients. However, most of the existing works still consider clients’ data distributions to be static or conforming to a simple dynamic, e.g., in participation rates of clients. In real FL applications, client data distributions change over time, and the dynamics, i.e., the evolving pattern, can be highly non-trivial. Furthermore, evolution may take place from training to testing. In this paper, we address dynamics in client data distributions and aim to train FL systems from time-evolving clients that can generalize to future target data. Specifically, we propose two algorithms, FedEvolve and FedEvp, which are able to capture the evolving patterns of the clients during training and are test-robust under evolving distribution shifts. FedEvolve explicitly models the temporal evolution by learning two distinct representation mappings that capture the transition between consecutive data domains for each client. In addition, FedEvp learns a single, evolving-domain-invariant representation by aligning current data with prototypes that are continuously updated from all previously seen domains. Through extensive experiments on both synthetic and real data, we show the proposed algorithms can significantly outperform the FL baselines across various network architectures.

Knowledge graphs play a crucial role in addressing the complexities of cybersecurity, as the increasing frequency and sophistication of cyber threats pose significant challenges to traditional defense technologies. In this paper, we propose a novel reasoning model, called INCYSER, that is tailored for cybersecurity. By leveraging Hyperdimensional Computing (HDC) as a symbolic and transparent computational model, INCYSER offers efficient and interpretable reasoning capabilities, ensuring reliable and trustworthy outcomes. Our model combines embedding-based unsupervised learning and HDC-based graph representation learning to construct a general representation for cybersecurity knowledge graphs, enabling diverse tasks including reasoning and general graph operations. Experimental evaluations demonstrate the effectiveness and efficiency of INCYSER, surpassing state-of-the-art models in link prediction and triple classification tasks. Additionally, a comprehensive ablation study examines the impact of various hyperparameters, showcasing the versatility of INCYSER. This work contributes to advancing the field of cybersecurity by introducing an interpretable and representation-based reasoning model for cybersecurity knowledge graphs.

The increasing popularity of virtual reality (VR) has stressed the importance of authenticating VR users while preserving their privacy. Behavioral biometrics, owing to their robustness and ease of collection, compared to traditional modes such as passwords, have become a favored authentication choice. While current approaches that utilize behavioral biometrics to train classifiers for authentication yield promising accuracy, they cause privacy breaches by sharing sensitive data with a server to train a central model. In this paper, we present MetaFL, a first-of-its-kind privacy-preserving VR authentication framework that leverages federated learning (FL) on multi-modal motion data. The design of MetaFL is motivated by our key insight that various modalities of motion data uniquely affect authentication performance for individual users and among different users. It is attributed to the fundamental challenge of privacy-preserving user authentication: users can access only their own data with limited global knowledge. To tackle this issue, MetaFL judiciously selects the most suitable modalities for each user, which is decomposed into within-user ordering and between-user selection to eliminate the complex interplay between various conflicting factors. Moreover, we develop a personalized strategy to initialize FL models, further improving authentication accuracy. Our extensive performance evaluation on six public datasets shows that MetaFL outperforms state-of-the-art FL-based models (e.g., 17--28% higher authentication accuracy), and its accuracy gap with the non-privacy-preserving central model is small (i.e., only <2%).

Heart rate is one of the most critical metrics for human health. Most common methods for measuring human heart rate involve body contact, whether from wearable devices or manual measurement. However, such devices can cause discomfort to some patients. Past work for non-contact or remote heart rate measurement (e.g., camera or radio) is often limited by line-of-sight requirements that are not always possible in the real-world environment. This paper presents FloHR, an indirect heart rate monitoring system for human beings using heartbeat-induced floor vibrations. The key insight is that the human body generates a small wave of pressure and sound with each heartbeat. These are propagated as vibration through the structures the person is in contact with (e.g., a chair) and through the floor. FloHR then detects and interprets these small floor vibrations. We developed a highly sensitive vibration sensing system and heartbeat pattern modelling to identify these tiny vibrations among other body motions and ambient noise. We evaluated FloHR in a real home environment, demonstrating an average heart rate error similar to medical device standards on the floor near the subjects' chair, and on the order of 10 beats per minute (bpm) on the floor 2 meters away from the subject.

Container security has received much research attention recently. Previous work has proposed to apply various machine learning techniques to detect security attacks in containerized applications. On one hand, supervised machine learning schemes require sufficient labeled training data to achieve good attack detection accuracy. On the other hand, unsupervised machine learning methods are more practical by avoiding training data labeling requirements, but they often suffer from high false alarm rates. In this article, we present a generic self-supervised hybrid learning (SHIL) framework for achieving efficient online security attack detection in containerized systems. SHIL can effectively combine both unsupervised and supervised learning algorithms but does not require any manual data labeling. We have implemented a prototype of SHIL and conducted experiments over 46 real-world security attacks in 29 commonly used server applications. Our experimental results show that SHIL can reduce false alarms by 33%–93% compared to existing supervised, unsupervised, or semi-supervised machine learning schemes while achieving a higher or similar detection rate.

Well-trained deep neural networks (DNNs) treat all test samples equally during prediction. Adaptive DNN inference with early exiting leverages the observation that some test examples can be easier to predict than others. This paper presents EENet, a novel early-exiting scheduling framework for multi-exit DNN models. Instead of having every sample go through all DNN layers during prediction, EENet learns an early exit scheduler, which can intelligently terminate the inference earlier for certain predictions, which the model has high confidence of early exit. As opposed to previous early-exiting solutions with heuristics-based methods, our EENet framework optimizes an early-exiting policy to maximize model accuracy while satisfying the given per-sample average inference budget. Extensive experiments are conducted on four computer vision datasets (CIFAR-10, CIFAR-100, ImageNet, Cityscapes) and two NLP datasets (SST-2, AgNews). The results demonstrate that the adaptive inference by EENet can outperform the representative existing early exit techniques. We also perform a detailed visualization analysis of the comparison results to interpret the benefits of EENet.