Federated Learning (FL) is a prominent distributed learning paradigm facilitating collaboration among nodes within an edge network to co-train a global model without centralizing data. By shifting computation to the network edge, FL offers robust and responsive edge-AI solutions and enhance privacy-preservation. However, deploying deep FL models within edge environments is often hindered by communication bottlenecks, data heterogeneity, and memory limitations. To address these challenges jointly, we introduce FeDEQ, a pioneering FL framework that effectively employs deep equilibrium learning and consensus optimization to exploit a compact shared data representation across edge nodes, allowing the derivation of personalized models specific to each node. We delve into a unique model structure composed of an equilibrium layer followed by traditional neural network layers. Here, the equilibrium layer functions as a global feature representation that edge nodes can adapt to personalize their local layers. Capitalizing on FeDEQ's compactness and representation power, we present a novel distributed algorithm rooted in the alternating direction method of multipliers (ADMM) consensus optimization and theoretically establish its convergence for smooth objectives. Experiments across various benchmarks demonstrate that FeDEQ achieves performance comparable to state-of-the-art personalized methods while employing models of up to 4 times smaller in communication size and 1.5 times lower memory footprint during training.

This project addresses the critical need for network-wide anomaly detection within the IoT landscape, a domain where devices frequently exhibit significant security vulnerabilities. We introduce an innovative solution that applies distributed computing and machine learning techniques to distinguish between normal and anomalous network traffic. To overcome the traditional constraints of machine learning, such as privacy issues and limited computational capacity on IoT devices, we propose a federated optimization framework, enabling a collaborative yet privacy-conscious profiling of network behavior across distributed IoT devices. The framework ensure rapid model training without incurring detection delays, even on devices with minimal processing power. Our methodology promises to set a new standard for efficient and privacy-aware anomaly detection in IoT networks.

In federated learning, participating clients typically possess non-i.i.d. data, posing a significant challenge to generalization to unseen distributions. To address this, we propose a Wasserstein distributionally robust optimization scheme called WAFL. Leveraging its duality, we frame WAFL as an empirical surrogate risk minimization problem, and solve it using a local SGD-based algorithm with convergence guarantees. We show that the robustness of WAFL is more general than related approaches, and the generalization bound is robust to all adversarial distributions inside the Wasserstein ball (ambiguity set). Since the center location and radius of the Wasserstein ball can be suitably modified, WAFL shows its applicability not only in robustness but also in domain adaptation. Through empirical evaluation, we demonstrate that WAFL generalizes better than the vanilla FedAvg in non-i.i.d. settings, and is more robust than other related methods in distribution shift settings. Further, using benchmark datasets we show that WAFL is capable of generalizing to unseen target domains. .

This project introduces a comprehensive air quality monitoring system tailored for smart cities, with a pioneering deployment in Ho Chi Minh City, one of Vietnam's largest urban centers. Our innovative system integrates direct measurements from strategically placed sensors with analytical data derived from remote sensing imagery. The core of the system lies in its sensor network, which provides precise measurements of key pollutants, allowing for an accurate calculation of the Air Quality Index (AQI). Complementing this, the analysis of remote sensing images offers a cost-effective method to approximate AQI over expansive areas not covered by sensors. These multifaceted data streams create a dynamic and robust tool for urban air quality management, promising enhanced environmental health and informed policy-making for smart cities.

Software-Defined Networking (SDN) is seen as a next-generation paradigm promising to build a vendor-neutral networking environment. By decoupling control plane from data plane, SDN shifts network intelligent logic into a logically centralized controller, thereby helping address many thorny problems in conventional network architecture. Despite of offering immense benefits, SDN has shown to be vulnerable to cyber attacks; meanwhile, Machine Learning (ML) has come into being the most powerful weapon to deal with those of security issues. In this paper, we proposed an improved solution of ML-based network intrusion detection system for better protecting SDN from malicious activities. The proposed solution is formed from a combination of ML techniques including Deep Sparse Autoencoder for reducing dimension and learning meaningful feature representation in network data; Conditional Generative Adversarial Network for solving data imbalance problem in intrusion detection datasets; and Ensemble Learning methods for classifying anomaly network traffic. Moreover, we leverage NetFPGA, a high-speed networking platform, to accelerate the packet processing task for the proposed system. By evaluating on empirical datasets, we show that our proposed system is capable of fast classification network traffic with high detection accuracy rate and relatively low false negative/positive rate.

This project employs a cutting-edge system utilizing IoT and cloud computing to deliver real-time, accurate monitoring of water salinity levels in Go Cong district, Tien Giang province, Vietnam. Through a network of IoT sensors, salinity data is continuously collected and then processed and analyzed using cloud technology, ensuring accessibility and scalability. This system not only provides immediate data access through a user-friendly dashboard but also features automated alerts for quick response to critical salinity fluctuations. With its advanced data analytics capabilities, the project is a significant step towards proactive water resource management, preserving the ecological balance and supporting the local communities in Go Cong.