This study deals with collaborative yet privacypreserving estimation algorithms in a sensor network. In particular, the sensing agents sample from Gaussian densities with heterogeneous private covariances, and aim to cooperatively compute the maximum likelihood estimate (MLE) of the common mean. To maintain accuracy and minimize computational load, we adopt the state decomposition (SD) approach that splits the estimated variable into private and public components. First, we design a privacy attack model (PAM) to show that the existing SD-based privacy-preserving algorithm for dynamic consensus does not remain simultaneously private and converge with convergent inputs. To address this issue, we modify the existing continuous-time distributed SD algorithm for discretetime networked updates with an additional initial mixing step and bounded inputs. When true covariances are known, the algorithm computes the mean MLE estimate while keeping the data and covariances private, and converges at the same rate as the non-private, centralized setting. When agents are estimating covariances locally, the algorithm estimates the mean MLE while keeping the local covariances private and maintains the same convergence rate as the local estimates. The rate compared to local estimates depends on the network structure. In simulations, the estimates from the proposed algorithms converge faster than the sample averaging and standard consensus methods.

Multimodal deep learning systems are deployed in dynamic scenarios due to the robustness afforded by multiple sensing modalities. Nevertheless, they struggle with varying compute resource availability (due to multi-tenancy, device heterogeneity, etc.) and fluctuating quality of inputs (from sensor feed corruption, environmental noise, etc.). Statically provisioned multimodal systems cannot adapt when compute resources change over time, while existing dynamic networks struggle with strict compute budgets. Additionally, both systems often neglect the impact of variations in modality quality. Consequently, modalities suffering substantial corruption may needlessly consume resources better allocated towards other modalities. We propose ADMN, a layer-wise Adaptive Depth Multimodal Network capable of tackling both challenges - it adjusts the total number of active layers across all modalities to meet compute resource constraints, and continually reallocates layers across input modalities according to their modality quality. Our evaluations showcase ADMN can match the accuracy of state-of-the-art networks while reducing up to 75% of their floating-point operations.

Complex events (CEs) play a crucial role in CPS-IoT applications, enabling high-level decision-making in domains such as smart monitoring and autonomous systems. However, most existing models focus on short-span perception tasks, lacking the long-term reasoning required for CE detection. CEs consist of sequences of short-time atomic events (AEs) governed by spatiotemporal dependencies. Detecting them is difficult due to long, noisy sensor data and the challenge of filtering out irrelevant AEs while capturing meaningful patterns. This work explores CE detection as a case study for CPS-IoT foundation models capable of long-term reasoning. We evaluate three approaches: (1) leveraging large language models (LLMs), (2) employing various neural architectures that learn CE rules from data, and (3) adopting a neurosymbolic approach that integrates neural models with symbolic engines embedding human knowledge. Our results show that the state-space model, Mamba, which belongs to the second category, outperforms all methods in accuracy and generalization to longer, unseen sensor traces. These findings suggest that state-space models could be a strong backbone for CPS-IoT foundation models for long-span reasoning tasks.

This work develops a distributed algorithm to enable cooperative tracking of the average of time-varying signals while preserving their privacy. We discuss the state decomposition approach, which incorporates public and private versions of estimates at each agent. In the presence of convergent inputs, we show that the existing state decomposition algorithm may not preserve privacy for the convergent input splits that allow convergence, and fail to converge for certain admissible splits. Therefore, we present a novel privacy-preserving algorithm that provides the same order of convergence rates on the estimates as that of the input signals. Further, the algorithm preserves privacy by letting a range of feasible input signals yield the same sequence of public variables. Finally, the algorithm is implemented for maximum likelihood estimation of the mean of data sampled by agents from Gaussians with private and unknown covariances.

This paper proposes U-FARE, an uncertainty-aware fair federated learning (FL) framework aimed at improving disease prediction in healthcare, with a specific focus on Alzheimer's disease detection. U-FARE incorporates evidential neural networks (ENN) to quantify uncertainty, enhancing both model fairness and accuracy. The framework ensures group-level fairness, providing consistent model performance across diverse healthcare environments despite data heterogeneity. We evaluate U-FARE on three real-world healthcare datasets-NACC, OASIS, and ADNI-comparing its performance to several state-of-the-art fairness-aware FL methods. Experimental results demonstrate that U-FARE outperforms baseline methods in both prediction accuracy and fairness, effectively balancing these two crucial aspects. The results also reveal the trade-off between fairness and accuracy, where higher fairness levels may compromise prediction accuracy. U-FARE achieves the highest accuracy (0.928) on the NACC dataset, consistently outperforms the competitive baseline q-FedAvg by 46%, particularly when higher fairness constraints are applied, and outperforms methods like Ditto and q-FFL with minimal accuracy variance and loss disparity. This is the first approach to simultaneously optimize fairness and accuracy in FL for Alzheimer's disease detection, providing a novel solution to the challenge of fair and effective AI in healthcare. The framework demonstrates the potential to address data heterogeneity while ensuring privacy and fairness in real-world applications.

Modern machine learning models excel at detecting individual actions, objects, or scene attributes from short, local observations. However, many real-world tasks, such as in smart cities and healthcare, require reasoning over complex events (CEs): (spatio)temporal, rule-governed patterns of short-term atomic events (AEs) that reflect high-level understanding and critical changes in the environment. These CEs are difficult to detect online: they are often rare, require long-range reasoning over noisy sensor data, must generalize rules beyond fixed-length traces, and suffer from limited real-world datasets due to the high annotation burden. We propose NAROCE, a Neural Algorithmic Reasoning framework for Online CE detection that separates the task into two stages: (i) learning CE rules from large-scale, low-cost pseudo AE concept traces generated by simulators or LLMs, and (ii) training an adapter to map real sensor data into the learned reasoning space using fewer labeled sensor samples. Experiments show that NAROCE outperforms the strongest baseline in accuracy, generalization to longer, unseen sequences, and data efficiency, achieving comparable performance with less than half the labeled data. These results suggest that decoupling CE rule learning from raw sensor inputs improves both data efficiency and robustness.

Advanced Persistent Threats (APTs) represent a significant challenge in cybersecurity due to their prolonged, multi-stage nature and the sophistication of their operators. Traditional detection systems typically focus on identifying malicious activity in binary terms (benign or malicious) without accounting for the progression of an attack. However, effective response strategies depend on accurate inference of the attack's current stage, as countermeasures must be tailored to whether an adversary is in the early reconnaissance phase or actively conducting exploitation or exfiltration. This work addresses the problem of attack stage inference under uncertainty, with a focus on robustness to out-of-distribution (OOD) inputs. We propose a classification approach based on Evidential Deep Learning (EDL), which models predictive uncertainty by outputting parameters of a Dirichlet distribution over possible stages. This allows the system not only to predict the most likely stage of an attack but also to indicate when it is uncertain or the input lies outside the training distribution. Preliminary experiments in a simulated environment demonstrate that the proposed model can accurately infer the stage of an attack with calibrated confidence while effectively detecting OOD inputs, which may indicate changes in the attackers' tactics. These results support the feasibility of deploying uncertainty-aware models for staged threat detection in dynamic and adversarial environments.

Federated Learning (FL) offers a promising approach for collaborative model training in healthcare while preserving data privacy. However, existing FL methods often fall short in addressing two critical challenges: client-level fairness and compounded uncertainty from data heterogeneity and privacy-preserving mechanisms. We propose fair-LDP, a fairness-aware Local Differential Privacy framework that promotes fairness and privacy via uncertainty-guided aggregation in federated healthcare AI. fair-LDP leverages evidential neural networks (ENNs) to quantify predictive uncertainty and introduces a novel strategy that uses uncertainty-driven local differential privacy to guide fairness-aware updates while preserving data privacy. This ensures equitable performance across clients with varying data quality while mitigating the influence of unreliable or outlier updates. fair-LDP incorporates an adaptive mechanism that adjusts each client's privacy budget based on model performance, balancing fairness, privacy, and accuracy. We evaluate fair-LDP on real-world healthcare datasets under both IID and non-IID settings. Our experimental results show that it consistently outperforms state-of-the-art fairness-aware and privacy-preserving FL baselines, with no added computational overhead, while maintaining privacy guarantees comparable to homomorphic encryption and secure multiparty computation. By integrating uncertainty modeling, fairness-aware aggregation, and adaptive local differential privacy, fair-LDP provides a practical and principled solution for responsible, equitable, and privacy-preserving federated learning in healthcare.

We study the problem of quickest change detection (QCD) when the post-change distribution has parametric uncertainty. The generalized likelihood ratio (GLR) cumulative sum (CuSum) procedure is known to be asymptotically optimum in this setting. However, this rule requires significant memory and computational resources, making it difficult to implement in practice. To overcome this limitation, sliding window approaches, such as the window-limited GLR CuSum and window-limited adaptive CuSum tests, have been employed, where the test statistic is computed over a fixed window of the latest observations. We propose the K-Best rule which instead keeps track of K hypothesized change points that have the largest test statistic. This allows the hypothesized change points to reduce epistemic uncertainty over time, while restricting the number of hypothesized change points considered. We characterize the growth rate of the K-Best window necessary to achieve the detection performance of the GLR-CuSum rule and quantify the computational benefits over the existing windowing approaches.

The Competitive Influence Maximization (CIM) problem involves entities competing to maximize influence in online social networks (OSNs). While Deep Reinforcement Learning (DRL) methods have shown promise, most assume binary user opinions and overlook behavioral factors. We introduce DRIM, a novel DRL-based CIM framework using Subjective Logic (SL) to incorporate user preferences and uncertainty, optimizing seed selection to spread true information while countering false information. DRIM’s Uncertainty-based Opinion Model (UOM) provides a realistic representation of user opinions. Results demonstrate that UOM maintains over 80% true influence against advanced misinformation, and DRIM outperforms state-of-the-art methods by up to 45% in influence and 77% in speed. DRIM also excels in limited-resource scenarios, networks with 10% invisibility, and when users are inclined to doubt true information.

Teams of embodied AI-enabled agents are critical for applications in extreme and highly dynamic environments. Developing robust controllers for such agents requires a deep understanding of the challenges encountered when attempting to coordinate and synchronize their individual perception-cognition-communication-action (PCCA) loops for team-wide mission objectives. We introduce a framework to explore the coordination of the PCCA loops across multiple agents in a new simulated physical environment designed to explore collaboration in each PCCA stage. This environment tasks teams of agents with the correct disposal of dangerous objects in an area and forces careful coordination of sensing, communication, movement, and manipulation actions by providing spatially-bounded communication, incorporating situations that require concerted effort by groups of agents, and introducing uncertainty into agents' sensing capabilities. We provide a set of heuristic controllers, an offline oracle model, and an initial exploration of a Reward Machine-based controller that learns its policies from training. Together these approaches serve to provide insights into the complexity of the multi-agent PCCA loop coordination problem. The multiagent PCCA simulation environment, which supports AI and human-controlled agents, and the code for various agent controllers are available at https://github.com/nesl/AI-Collab.

An in-depth understanding of uncertainty is the first step to making effective decisions under uncertainty. Machine/deep learning (ML/DL) has been hugely leveraged to solve complex problems involved with processing high-dimensional data. However, reasoning and quantifying different uncertainties to achieve effective decision-making have been much less explored in ML/DL than in other Artificial Intelligence (AI) domains. In particular, belief/evidence theories have been studied in Knowledge representation and reasoning (KRR) since the 1960s to reason and measure uncertainties to enhance decision-making effectiveness. Based on our in-depth literature review, only a few studies have leveraged mature uncertainty research in belief/evidence theories in ML/DL to tackle complex problems under different types of uncertainty. Our present survey paper discusses major belief theories and their core ideas dealing with uncertainty causes and types and quantifying them, along with the discussions of their applicability in ML/DL. Particularly, we discuss three main approaches leveraging belief theories in Deep Neural Networks (DNNs), including Evidential DNNs, Fuzzy DNNs, and Rough DNNs, in terms of their uncertainty causes, types, and quantification methods along with their applicability in diverse problem domains. Through an in-depth understanding of the extensive survey on this topic, we discuss insights, lessons learned, limitations of the current state-of-the-art bridging belief theories and ML/DL, and future research directions. This paper conducts an extensive survey by bridging belief theories and deep learning in reasoning and quantifying uncertainty to help researchers initiate uncertainty and decision-making research.

Competitive Influence Maximization (CIM) involves entities competing to maximize influence in online social networks (OSNs). Current Deep Reinforcement Learning (DRL) methods in CIM rely on simplistic binary opinion models (i.e., an opinion is represented by either 0 or 1) and often overlook the complexity of user behaviors and prior knowledge. We propose a novel DRL-based framework that enhances CIM analysis by integrating Subjective Logic (SL) to accommodate uncertain opinions, user behaviors, and preferences. This approach targets the mitigation of false information by effectively propagating true information. By modeling two competitive agents, one spreading true information and the other spreading false information, we capture the strategic interplay essential to CIM. Our framework utilizes an uncertainty-based opinion model (UOM) to assess the impact on information quality in OSNs, emphasizing the importance of user behavior alongside network topology in selecting influential seed nodes. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art methods, achieving faster and more influential results (i.e., outperforming over 20 method shows robust performance in partially observable networks, effectively doubling the performance when users are predisposed to disbelieve true information.

Human decision-making, as well as control of autonomous systems, have deployed deep learning models for detecting complex events from unstructured sensory data. However, the strong performance of these models is restricted to events with short intervals of time and space due to the limited context memory of their architectures. Thus, detecting events that transpire over long periods of time with multiple spatially distant sensor sources (known as complex events) remains challenging for these purely neural-based methods, particularly as environmental conditions and object appearances change. In recent years, neurosymbolic approaches have been proposed that use both neural-based perception and symbolic reasoning for capturing complex events. However, these approaches still face issues of adaptation to perceptual domain shift in complex events. We address these problems in the context of a prototype neurosymbolic system called DANCER, which performs Domain Adaptation and Neurosymbolic inference in Complex Event Reasoning. DANCER aims to provide domain adaptation in a post-deployment setting while minimizing runtime user burden for annotation. To enable training and evaluation of DANCER, we also provide a physics-based synthetic sensor data generator to create videos given complex scenario specifications. We evaluate DANCER on a dataset of generated synthetic data. We show that DANCER yields a 48% increase in accuracy of complex event detection using domain adaptation while significantly reducing the annotation time of our synthetic complex events by up to 2.7x, demonstrating DANCER's ability to effectively detect complex events under perceptual domain shift.

Steerable cameras that can be controlled via a network, to retrieve telemetries of interest have become popular. In this paper, we develop a framework called AcTrak, to automate a camera’s motion to appropriately switch between (a) zoom ins on existing targets in a scene to track their activities, and (b) zoom out to search for new targets arriving to the area of interest. Specifically, we seek to achieve a good trade-off between the two tasks, i.e., we want to ensure that new targets are observed by the camera before they leave the scene, while also zooming in on existing targets frequently enough to monitor their activities. There exist prior control algorithms for steering cameras to optimize certain objectives; however, to the best of our knowledge, none have considered this problem, and do not perform well when target activity tracking is required. AcTrak automatically controls the camera’s PTZ configurations using reinforcement learning (RL), to select the best camera position given the current state. Via simulations using real datasets, we show that AcTrak detects newly arriving targets 30% faster than a non-adaptive baseline and rarely misses targets, unlike the baseline which can miss up to 5% of the targets. We also implement AcTrak to control a real camera and demonstrate that in comparison with the baseline, it acquires about 2× more high resolution images of targets.

Network traffic data analysis is important for securing our computing environment and data. However, analyzing network traffic data requires tremendous effort because of the complexity of continuously changing network traffic patterns. To assist the user in better understanding and analyzing the network traffic data, an interactive web-based visualization system is designed using multiple coordinated views, supporting a rich set of user interactions. For advancing the capability of analyzing network traffic data, feature extraction is considered along with uncertainty quantification to help the user make precise analyses. The system allows the user to perform a continuous visual analysis by requesting incrementally new subsets of data with updated visual representation. Case studies have been performed to determine the effectiveness of the system. The results from the case studies support that the system is well designed to understand network traffic data by identifying abnormal network traffic patterns.

Detecting complex events from subsymbolic data streams (such as images, audio recordings or videos) is a challenging problem, as traditional symbolic approaches cannot be used to process subsymbolic data, and neural-only approaches usually require larger amounts of training data than available. In this paper, we present DeepProbCEP, a Complex Event Processing (CEP) approach designed with four objectives: (i) allowing the use of subsymbolic data as an input, (ii) retaining flexibility and modularity in the definition of complex event rules, (iii) limiting the cost of obtaining training data and (iv) being robust against adversarial conditions. DeepProbCEP archives this by using a neuro-symbolic approach, which combines the neural and symbolic approaches to allow training with sparse data. This is made possible through the injection of human knowledge. In this paper, we demonstrate that DeepProbCEP outperforms other state-of-the-art approaches when training using sparse data. We also show that DeepProbCEP is robust in different adversarial settings. Finally, DeepProbCEP’s flexibility is demonstrated by showing it can be used to process both images and audio as input.