Balochistan University of Engineering and Technology

The spread of deepfakes poses significant security concerns, demanding reliable detection methods. However, diverse generation techniques and class imbalance in datasets create challenges. We propose CAE-Net, a Convolution- and Attention-based weighted Ensemble network combining spatial and frequency-domain features for effective deepfake detection. The architecture integrates EfficientNet, Data-Efficient Image Transformer (DeiT), and ConvNeXt with wavelet features to learn complementary representations. We evaluated CAE-Net on the diverse IEEE Signal Processing Cup 2025 (DF-Wild Cup) dataset, which has a 5:1 fake-to-real class imbalance. To address this, we introduce a multistage disjoint-subset training strategy, sequentially training the model on non-overlapping subsets of the fake class while retaining knowledge across stages. Our approach achieved 94.46% accuracy and a 97.60% AUC, outperforming conventional class-balancing methods. Visualizations confirm the network focuses on meaningful facial regions, and our ensemble design demonstrates robustness against adversarial attacks, positioning CAE-Net as a dependable and generalized deepfake detection framework.

Urban land-use allocation represents a complex multi-objective optimization problem critical for sustainable urban development policy. This paper presents novel computational intelligence approaches for optimizing land-use allocation in mixed-use areas, addressing inherent trade-offs between land-use compatibility and economic objectives. We develop multiple optimization algorithms, including custom variants integrating differential evolution with multi-objective genetic algorithms. Key contributions include: (1) CR+DES algorithm leveraging scaled difference vectors for enhanced exploration, (2) systematic constraint relaxation strategy improving solution quality while maintaining feasibility, and (3) statistical validation using Kruskal-Wallis tests with compact letter displays. Applied to a real-world case study with 1290 plots, CR+DES achieves 3.16% improvement in land-use compatibility compared to state-of-the-art methods, while MSBX+MO excels in price optimization with 3.3% improvement. Statistical analysis confirms that algorithms incorporating difference vectors significantly outperform traditional approaches across multiple metrics. The constraint relaxation technique enables broader solution space exploration while maintaining practical constraints. These findings provide urban planners and policymakers with evidence-based computational tools for balancing competing objectives in land-use allocation, supporting more effective urban development policies in rapidly urbanizing regions.

This study quantifies the impact of projected climate change on the daily streamflow of the Brahmaputra River (Bahadurabad outlet, Bangladesh) using four data-driven models: support vector machine (SVM), random forest (RF), long short-term memory (LSTM), and bidirectional LSTM (Bi-LSTM). Trained on 1981-2008 data with rainfall and temperature as predictors and tested on 2009-2014 (Bi-LSTM and RF outperformed with R-2 approximate to 0.90), the models were forced with bias-corrected projections from 13 CMIP6 GCMs under six composite scenarios (coolest to wettest) for the 2030s, 2050s, and 2080s. Key findings indicate a significant alteration of the hydrograph, characterized by an earlier monsoon rise, higher July-August flow plateaus (similar to 50,000-70,000 m(3)/s by the 2080s under warm/wet scenarios), and a slower recession. Projections also show intensifying extremes, with median monsoon monthly maxima reaching similar to 66,000-90,000 m(3)/s by the 2080s, model-dependent. Crucially, the analysis reveals a trend toward greater seasonal variability, where the wet season becomes wetter while the dry season may become even drier, particularly under the driest and coolest scenarios. Mean annual flow changes by the 2080s range from -1% (coolest) to +50% (wettest). The results unanimously project stronger, longer monsoon flows and amplified peaks, and substantially extended flood risk, particularly under warmer and wetter futures. [GRAPHICS]

Achieving uniform melt-pool geometry in laser powder bed fusion (LPBF) remains difficult due to nonlinear parameter interactions, stochastic powder-bed conditions, and progressive heat accumulation. These challenges are intensified by the lack of a sustainable digital manufacturing platform that can unify data, models, and control. To address this issue, this work introduces a unified LLM-agent-guided framework that integrates stochastic calibration and real-time control within a single pipeline. The work uses two co-ordinated agents. CalibAgent performs Bayesian Markov Chain Monte Carlo calibration of an analytical thermal model of a multi-layer, and multi-track LPBF process using NIST AM-Bench meltpool data, inferring distributions over absorptivity and thermal diffusivity. ControlAgent then uses the calibrated model to compute spatially varying laser-power schedules via analytical inversion, requiring no iterative optimization and executing in milliseconds. Demonstrations on IN718 across four scan geometries (horizontal and 45° cross-hatch, concentric triangular hatch, and a multi-layer L-shaped build) show that the controller reduces steady-state depth fluctuations by 83%–99% and eliminates turnaround spikes of up to 41 μm. In the multi-layer case, it maintains near-uniform depth across four layers without retraining or parameter adjustment. The framework requires no task-specific training, reward engineering, or expert tuning, enabling immediate extension to new materials and scan strategies.

Deepfake detectors are almost always deployed as the single checkpoint that scored best on a validation split, yet the reliability of that choice is rarely measured. We measure it with PairCLIP-SWA, a single-backbone detector that adapts an OpenCLIP ViT-L/14 visual tower by progressive block unfreezing on a class-balanced corpus built by CLIP-similarity-guided one-to-one subsampling, and averages the fine-tuning trajectory in weight space. Trained in-domain on the IEEE SP Cup 2025 (DF-Wild-Cup) benchmark under the same protocol as the strongest published three-backbone ensemble, the weight-averaged detector reaches 97.04