Interdependent infrastructures rarely fail in isolation: outages in one network can cascade across layer boundaries and hinder restoration elsewhere. This study links three real-world data sets: an IEEE 118-bus power grid, a CAIDA AS-level Internet snapshot, and the EPANET Net-3 water model into a sparse but asymmetric three-layer “supra-graph.” A hybrid connectivity-and-load failure (CF) rule and a dynamic-budget self-healing (SH) routine then compete on that multiplex. The topological-nearest coupling scheme more than doubles the density of water-to-communication links relative to other off-diagonal blocks, creating a handful of cross-layer hubs that carry 72 percent of all inter-layer traffic. Simulation results show that these hubs deepen the initial cascade dip and prolong recovery: the active-node fraction drops to 0.86 within five steps yet requires roughly twenty steps of healing to regain full functionality. A multi-objective genetic algorithm explores trade-offs among healing budget β, connectivity threshold ϕ_c, and overload threshold ρ_c. Sixteen Pareto-optimal solutions trace a concave frontier: raising β from 0.20 to 0.35 yields the same resilience gain as tightening ϕ_c from 0.60 to 0.75, while further increases in either lever give diminishing returns. A balanced setting (β ≈ 0.30, ϕ_c ≈ 0.65, ρ_c ≈ 1.0) restores 80 percent of nodes at minimum cost. The findings highlight how cross-layer asymmetry skews both failure propagation and optimal resource allocation, offering infrastructure operators a quantified basis for sizing repair crews and setting connectivity targets under interdependent risk.

Cascading link failures continue to imperil power grids, transport networks, and cyber-physical systems, yet the relationship between a network’s robustness at the moment of attack and its subsequent resiliency remains poorly understood. We introduce a dynamic framework in which connectivity-based cascades and distributed self-healing act concurrently within each time-step. Failure is triggered when a node’s active-neighbor ratio falls below a threshold $\phi$; healing activates once the global fraction of inactive nodes exceeds trigger $T$ and is limited by budget $B$. Two real data sets—a 332-node U.S. airport graph and a 1133-node university e-mail graph—serve as testbeds. For each graph we sweep the parameter quartet ($\phi ,\hspace{0.17em}B,\hspace{0.17em}T,attackmode$) and record (i) immediate robustness $R$, (ii) 90% recovery time T₉₀, and (iii) cumulative average damage. Results show that targeted hub removal is up to three times more damaging than random failure, but that prompt healing with $B\ge 0.12$ can halve T₉₀. Scatter-plot analysis reveals a non-monotonic correlation: high-$R$ states recover quickly only when $B$ and $T$ are favorable, whereas low-$R$ states can rebound rapidly under ample budgets. A multiplicative fit $T_{90}\propto B^{-\beta }g\left(T\right)h\left(R\right)$ (with $\beta \approx 1$) captures these interactions. The findings demonstrate that structural hardening alone cannot guarantee fast recovery; resource-aware, early-triggered self-healing is the decisive factor. The proposed model and data-driven insights provide a quantitative basis for designing infrastructure that is both robust to failure and resilient in restoration.

Smartphones are now deeply embedded in daily life, but excessive dependence may increase the risk of nomophobia, which is associated with anxiety, sleep disruption, and reduced productivity. Existing screening methods mainly rely on self-reported questionnaires, which are subjective and difficult to scale for continuous monitoring. This study proposes a privacy-preserving federated deep learning framework for three-level nomophobia risk prediction (Normal, Mild, and Severe) using smartphone usage logs while keeping raw user data on local devices. The proposed pipeline uses a publicly available secondary dataset with 1000 original records and expands it to 100,000 records through constraint-aware synthetic augmentation. A continuous risk score is computed from standardized smartphone usage indicators and then converted into three classes using tertile-based thresholds. Several local architectures, including CNN, MLP, ResMLP, Wide & Deep, and a lightweight TabNet-style gated model, are evaluated under FedAvg. In the reported experiments, differential privacy is enabled through DP-SGD with gradient clipping and Gaussian noise. To protect update transmission, the framework applies protected update sharing through encrypted transport of masked updates. Each client masks its local update and encrypts the masked payload before transmission. This mechanism improves communication confidentiality and reduces the direct exposure of client updates. Under a fixed federated setup with five clients and 25 communication rounds, tabular models achieved near-ceiling performance on the constructed test set. The MLP achieved 99.12% accuracy, 99.12% F1-score, 0.9868 MCC, and 0.9997 AUC, while Wide & Deep achieved 98.95% accuracy, 98.95% F1-score, 0.9843 MCC, and 0.9997 AUC. In contrast, sequential models such as RNN and LSTM showed near-random performance, suggesting that the current aggregated feature representation is better suited to tabular learning than temporal modeling. These results indicate that the proposed federated pipeline can effectively learn the constructed nomophobia risk labels while preserving local data ownership. However, because the labels are derived from usage features rather than clinical or psychometric assessment, the findings should be interpreted as proof-of-concept results for constructed risk labels rather than evidence of clinical diagnostic validity.