This study, set within the vast and varied research field of industrial Internet of Things (IoT) systems, proposes a methodology to address uncertainty quantification (UQ) issues in predictive maintenance (PdM) practices. At its core, this paper leverages the commercial modular aero-propulsion system simulation (CMAPSS) dataset to evaluate different artificial intelligence (AI) prognostic algorithms for remaining useful life (RUL) forecasting while supporting the estimation of a robust confidence interval (CI). The methodology primarily involves the comparison of statistical learning (SL), machine learning (ML), and deep learning (DL) techniques for each different scenario of the CMAPSS, evaluating the performances through a tailored metric, the S-score metric, and then benchmarking diverse conformal-based uncertainty estimation techniques, remarkably naive, weighted, and bootstrapping, offering a more suitable and reliable alternative to classical RUL prediction. The results obtained highlight the peculiarities and benefits of the conformal approach, despite probabilistic models favoring the adoption of complex models in cases where the operating conditions of the machine are multiple, and suggest the use of weighted conformal practices in non-exchangeability conditions while recommending bootstrapping alternatives for contexts with a more substantial presence of noise in the data.
9 June 2026 in ACM Transactions on Multimedia Computing, Communications, and Applications
This study presents a blockchain-based framework for user-generated content (UGC) validation. The methodology combines a bootstrapped holographic watermarking procedure with temporally distributed payloads, SHA-256 hashing linked to an Ethereum-compatible ledger, and a lightweight custom 3D CNN model for watermark detection. The blockchain layer functions as an audit trail, confirming ownership and integrity of the content. Meanwhile, watermark detection through the Model-as-a-Key (MaaK) paradigm provides a complementary mechanism for identifying embedded signatures when content alterations occur. To evaluate the proposed framework, we define intentional and non-intentional threat models on a subset of the Kinetics dataset, containing user-generated videos. Considering a mixed clean–noisy evaluation set, the custom watermark detector achieves an overall accuracy of \(91\%\) with an ROC–AUC of \(0.96\) , while presenting a mean accuracy of \(81\%\) under advanced threat conditions, indicating stable performance across multiple transformations. The ablation study, which compares our model with the pretrained VGG-16/19 and ResNet-50, highlights improved efficiency–robustness trade-offs within the integrated verification pipeline. In conclusion, we demonstrate the feasibility of combining blockchain with AI-based watermark recognition for supporting user-generated content authenticity verification in decentralized environments.
Semantic segmentation of crops and weeds is an important component of selective weeding systems, but accurate segmentation networks can impose substantial computational and memory demands on resource-constrained agricultural platforms. This paper presents a systematic study of the accuracy–efficiency trade-offs involved in compressing semantic segmentation networks for agricultural edge computing. Six architectures spanning conventional CNN encoder–decoders, atrous and pyramid-pooling decoders, a mobile-oriented CNN, and two transformer scales are evaluated on the Crop/Weed Field Image Dataset (CWFID), with three random seeds per architecture. The most favourable compact model is subsequently evaluated under unstructured global magnitude pruning and dependency-aware structured channel pruning, using one-shot and iterative prune-and-fine-tune schedules at sparsity levels from 10% to 80%, with three seeds per configuration. Among the dense baselines, U-Net with a MobileNetV3-Large encoder matches the accuracy of the larger SegFormer-B2 model (88.85% versus 88.67% mIoU, a difference of 0.18% across three seeds, p = 0.58) with 3.7× fewer parameters than SegFormer-B2. Iterative unstructured pruning reaches 80% weight sparsity while retaining 89.04% mIoU; however, because the dense tensor dimensions are unchanged, this sparsity does not reduce the parameter count, GFLOPs, or model size on conventional hardware. Structured channel pruning, which does reduce all three, remains viable only up to 30% sparsity, reducing GFLOPs by 2.1× at a cost of 4.21 mIoU points and collapsing beyond 40%. Without fine-tuning, structured pruning causes severe degradation even at 10% channel sparsity. The results show that architecture selection has a major influence on the accuracy–efficiency trade-off, while the effects of pruning depend strongly on the pruning structure and fine-tuning strategy. Crop segmentation consistently degrades before the aggregate mIoU indicates severe failure, highlighting the need for class-wise evaluation when compressing crop–weed segmentation models.
This paper presents a Retrieval Augmented Intrusion Detection system (RAID), a three-stage framework for interpretable AI-driven intrusion detection in Industrial Control Systems (ICS). Existing ICS detectors flag anomalies without explaining them, and LLM-based explanations are rarely verified against the detector’s rules or linked to past incidents. Specifically, a resilient isolation forest algorithm is used for its ability to generate structured anomaly interpretations and to mitigate the effects of dangerous data drift. A grounded physics-aware feature engineering process extends the baseline capabilities of this algorithm, including counterfactual evaluation, contrastive analysis, causal chain identification, and redundancy inspection of data streams. Subsequently, a fleet of Large Language Models (LLMs) is used to generate deep natural-language explanations for each anomaly log, under the guidance of an independent LLM orchestrator that produces a final incident report. Finally, a Retrieval-Augmented Generation framework embeds the historical patterns of the recovered events into an accessible index that contains all the log interpretations over time. The entire framework is validated at each stage on the HAI-21.03 dataset: among nine unsupervised detectors, the Isolation Forest provides the most balanced performance (F1 = 0.409, MCC = 0.393, eTaPR F1 = 0.773); 87% of the propositions generated by the language agents exactly match the underlying detection rules; and the retrieval stage reaches 0.998 faithfulness and 0.782 answer correctness.