Indian Institute of Technology Gandhinagar

Intrusion detection is essential for the security of modern networks against cyberattacks, but classical machine learning models like Support Vector Machines (SVM) often face challenges with high-dimensional data and scalability. In this work, we develop and explore the use of Quantum Support Vector Machines (QSVM) for intrusion detection using quantum kernels and feature maps to enhance classification accuracy. In this work, we compare the SVM and QSVM across four benchmark datasets, namely, CICIDS2017, NSL-KDD, CIDDS-001, and UGR’16 and also analyze the impact of applying Principal Component Analysis (PCA) for dimensionality reduction.Our results indicate that QSVM performs better in lower-dimensional spaces: on the CIDDS-001 dataset, QSVM achieved 97.1% accuracy compared to 94.5% for SVM, with a 3.2% increase in F1-score and 40% fewer false alarms. On the CICIDS2017 dataset, QSVM improved precision from 91% to 98% and reduced false positives by 98%, increasing overall reliability. Similarly, on the NSL-KDD dataset, QSVM achieved up to 1.5% higher accuracy and better F1-scores at moderate PCA dimensions (8-16) but lagged slightly in the lowest-dimensional case, suggesting its sensitivity to feature compression. However, for datasets with high dimensions like UGR’16, classical SVM maintained better performance, which shows the current limitations of QSVM scalability. Our study provides clear evidence that quantum-enhanced models can outperform classical SVMs under specific conditions, especially with reduced feature spaces.

Clay-rich soils are often used for constructing various compacted geotechnical structures. However, prolonged exposure to high temperatures makes them prone to desiccation cracking, compromising their structural integrity. This study investigated the desiccation crack formation in high plasticity compacted clay under controlled drying. An automated image acquisition system was used to monitor crack development. Digital Image Analysis (DIA) and Digital Image Correlation (DIC) were utilized to quantify crack morphology and strain localization. DIA results showed that crack area increased steadily after reaching a critical water content, followed by a decline due to the soil’s self-healing property. The crack network stabilized once the final critical water content was reached. DIC analysis revealed that cracks originated at high tensile strain zones. Crack propagation modes including opening, sliding, and mixed modes were observed. A wide primary crack subdivided the surface into two clods, each with a distinct shrinkage center, indicating low tessellation tendency.

Individual fairness guarantees are often desirable properties to have, but they become hard to formalize when the dataset contains outliers. Here, we investigate the problem of developing an individually fair k-median and k-means clustering algorithm for datasets that contain outliers. That is, given n points and k centers, we want that for each point which is not an outlier, there must be a center within the nk nearest neighbours of the given point. While a few of the recent works have looked into individually fair clustering, this is the first work that explores this problem in the presence of outliers for k-clustering. For this purpose, we define and solve a linear program (LP) that helps us identify the outliers. We exclude these outliers from the dataset and apply a rounding algorithm that computes the k centers, such that the fairness constraint of the remaining points is satisfied. We also provide theoretical guarantees that our method leads to a guaranteed approximation of the fair radius as well as the clustering cost. We also demonstrate our techniques empirically on real-world datasets.

5G Core (5GC) network functions (NFs), including the Access and Mobility Management Function (AMF), Session Management Function (SMF), Network Repository Function (NRF), and User Plane Function (UPF), are increasingly deployed on commodity server hardware. As these functions run on general-purpose servers, their performance is significantly influenced by underlying micro-architectural behaviors such as Cache and memory-access behavior. Understanding this behavior is therefore important for performance analysis in practical deployments. Most existing studies examine open-source 5GC implementations, however, prominent 5G cores incorporate vendor-specific optimizations in scheduling, memory management, and packet processing that can affect how they use CPU and memory resources. As a result, observations from open-source systems do not generalize to production grade deployments. In this work, we present a measurement-driven characterization of key commercial 5GC network functions deployed as a virtualized network function. Using hardware performance counters collected via the Linux perf, we analyze the IPC and cache-related behavior of AMF, SMF, NRF, and UPF under two operating conditions: idle operation without active user equipment and live traffic generated by UEs. Our analysis reveals micro-architectural characteristics across control-plane and user-plane functions and shows how traffic conditions redistribute execution and memory-system pressure across the 5GC. Overall, this study demonstrates that hardware performance counter–based profiling provides a practical and non-intrusive means to characterize commercial 5GC network functions under realistic deployment conditions.

In this paper, we present a Fast Motion Deblurring-Conditional Generative Adversarial Network (FMD-cGAN) that helps in blind motion deblurring of a single image. FMD-cGAN delivers impressive structural similarity and visual appearance after deblurring an image. Like other deep neural network architectures, GANs also suffer from large model size (parameters) and computations. It is not easy to deploy the model on resource constraint devices such as mobile and robotics. With the help of MobileNet based architecture that consists of depthwise separable convolution, we reduce the model size and inference time, without losing the quality of the images. More specifically, we reduce the model size by 3-60x compare to the nearest competitor. The resulting compressed Deblurring cGAN faster than its closest competitors and even qualitative and quantitative results outperform various recently proposed state-of-the-art blind motion deblurring models. We can also use our model for real-time image deblurring tasks. The current experiment on the standard datasets shows the effectiveness of the proposed method.