arXiv subjects

Sarbani Palit

Publications and source records attributed to Sarbani Palit.

8 recordsLinked to original sources

arXiv2026

TADP-RME: A Trust-Adaptive Differential Privacy Framework for Enhancing Reliability of Data-Driven Systems

Ensuring reliability in adversarial settings necessitates treating privacy as a foundational component of data-driven systems. While differential privacy and cryptographic protocols offer strong guarantees, existing schemes rely on a fixed privacy budget, leading to a rigid utility-privacy trade-off that fails under heterogeneous user trust. Moreover, noise-only differential privacy preserves geometric structure, which inference attacks exploit, causing privacy leakage. We propose TADP-RME (Trust-Adaptive Differential Privacy with Reverse Manifold Embedding), a framework that enhances reliability under varying levels of user trust. It introduces an inverse trust score in the range [0,1] to adaptively modulate the privacy budget, enabling smooth transitions between utility and privacy. Additionally, Reverse Manifold Embedding applies a nonlinear transformation to disrupt local geometric relationships while preserving formal differential privacy guarantees through post-processing. Theoretical and empirical results demonstrate improved privacy-utility trade-offs, reducing attack success rates by up to 3.1 percent without significant utility degradation. The framework consistently outperforms existing methods against inference attacks, providing a unified approach for reliable learning in adversarial environments.

cs.CR↗

arXiv2026

Blackening Cryosphere: Revealing Hotspot Shifts and HGB-Based Forecasting of Absorbing Aerosol Threats over the Himalayan Frozen Frontiers

Black carbon and mineral dust are key absorbing aerosols that influence atmospheric radiation and increasingly threaten global cryospheric stability. This study examines the long-range transport and seasonal variability of these aerosols over Pakistan and their movement toward the western Himalayas. Using satellite-derived Absorption Aerosol Optical Depth (AAOD) data from 2019 to mid-2025, we analyse their spatiotemporal behaviour across Pakistan's urban lowlands and high-altitude regions. Fifteen-day aggregated AAOD fields are used to track seasonal transport into glaciated terrain, where deposited aerosols can darken snow and ice and accelerate melt. For high-AAOD events, a probabilistic forecasting approach based on machine learning (ML) was developed. Using geographical, seasonal, and lagged indicators, a histogram-based gradient boosting classifier was trained to predict AAOD exceedance one step in advance. ROC-AUC, PR-AUC, and the Brier score were used to assess the model's performance. The results show high predictive capacity and good probability calibration, with values of 0.791, 0.269, and 0.028, respectively. Forecasts indicate that areas adjacent to Himalayan glaciers consistently exhibit the highest probability of increasing AAOD, signalling an elevated risk of aerosol-induced snowmelt.

physics.ao-ph↗

arXiv2025

GraMFedDHAR: Graph Based Multimodal Differentially Private Federated HAR

Human Activity Recognition (HAR) using multimodal sensor data remains challenging due to noisy or incomplete measurements, scarcity of labeled examples, and privacy concerns. Traditional centralized deep learning approaches are often constrained by infrastructure availability, network latency, and data sharing restrictions. While federated learning (FL) addresses privacy by training models locally and sharing only model parameters, it still has to tackle issues arising from the use of heterogeneous multimodal data and differential privacy requirements. In this article, a Graph-based Multimodal Federated Learning framework, GraMFedDHAR, is proposed for HAR tasks. Diverse sensor streams such as a pressure mat, depth camera, and multiple accelerometers are modeled as modality-specific graphs, processed through residual Graph Convolutional Neural Networks (GCNs), and fused via attention-based weighting rather than simple concatenation. The fused embeddings enable robust activity classification, while differential privacy safeguards data during federated aggregation. Experimental results show that the proposed MultiModalGCN model outperforms the baseline MultiModalFFN, with up to 2 percent higher accuracy in non-DP settings in both centralized and federated paradigms. More importantly, significant improvements are observed under differential privacy constraints: MultiModalGCN consistently surpasses MultiModalFFN, with performance gaps ranging from 7 to 13 percent depending on the privacy budget and setting. These results highlight the robustness of graph-based modeling in multimodal learning, where GNNs prove more resilient to the performance degradation introduced by DP noise.

cs.LG↗

arXiv2024

Quantized Redshift and its significance for recent observations

With the recent observational evidence in extra galactic astronomy, the interpretation of the nature of quasar redshift continues to be a research interest. Very high redshifts are being detected for extragalactic objects that are presumably very distant and young while also exhibiting properties that are characteristic of a more mature galaxy such as ours. According to Halton Arp and Geoffrey Burbidge, redshift disparities consist of an intrinsic component and are related to an evolutionary process. Karlsson observed redshift periodicity at integer multiples of 0.089 in log scale and Burbidge observed redshift periodicity at integer multiples of 0.061 in linear scale. Since Singular Value Decomposition (SVD) based periodicity estimation is known to be superior for noisy data sets, especially when the data contains multiple harmonics and overtones, mainly irregular in nature, we have chosen it to be our primary tool for analysis of the quasar-galaxy pair redshift data. We have observed a fundamental periodicity of 0.051 with a confidence interval of 95% in linear scale with the site-available Sloan Digital Sky Survey data release 7 (SDSS DR7) quasar-galaxy pair data set. We have independently generated quasar-galaxy pair data sets from both 2dF and SDSS and found fundamental periodicities of 0.077 and 0.089 in log scale with a confidence interval of 95%.

astro-ph.GA↗

arXiv2022

Redshift periodicity and its significance for Recent observation

Recent observational evidence in extra galactic astronomy, the interpretation of the nature of quasar redshift continues to be research interest. Spectrum observation of high redshift quasar is young in nature. Observational evidence discuss on physical interpretation of redshift periodicity with statistical confirmation. Karlsson observed redshift periodicity at integer multiples of 0.089 in log scale and Burbidge observed redshift periodicity integer multiple of 0.061 in linear scale .Data analysis is important in order to form correct interpretations of the observed phenomena. Since Singular value decomposition (SVD) based periodicity estimation is known to be superior for noisy data sets, especially when the data contains multiple harmonics and overtones, mainly irregular in nature, we have chosen it to be our primary tool for analysis of the quasar-galaxy pair redshift data. Kernel density estimation has been performed for estimating the bin width as proper computation of this quantity is crucial for the correctness of the analysis and prevention of over smoothing of the data.We observed fundamental periodicity to be an integer multiple of 0.063 and 0.0604 using method1 and method2 in the transformed quasar redshift data with 95% confidence interval in linear scale. Our results clearly establish that redshift is quantized for quasar-galaxy pair data and its histogram exhibits periodic peak(s). At last briefly discussed on physical interpretation of quantized redshift for quasar and galaxy.Hoyle Narlikar theory of gravity explain the Mystery in recent observation.

astro-ph.CO↗

arXiv2021

Integrating Unsupervised Clustering and Label-specific Oversampling to Tackle Imbalanced Multi-label Data

There is often a mixture of very frequent labels and very infrequent labels in multi-label datatsets. This variation in label frequency, a type class imbalance, creates a significant challenge for building efficient multi-label classification algorithms. In this paper, we tackle this problem by proposing a minority class oversampling scheme, UCLSO, which integrates Unsupervised Clustering and Label-Specific data Oversampling. Clustering is performed to find out the key distinct and locally connected regions of a multi-label dataset (irrespective of the label information). Next, for each label, we explore the distributions of minority points in the cluster sets. Only the minority points within a cluster are used to generate the synthetic minority points that are used for oversampling. Even though the cluster set is the same across all labels, the distributions of the synthetic minority points will vary across the labels. The training dataset is augmented with the set of label-specific synthetic minority points, and classifiers are trained to predict the relevance of each label independently. Experiments using 12 multi-label datasets and several multi-label algorithms show that the proposed method performed very well compared to the other competing algorithms.

cs.LG↗

arXiv2019

Fast Periodicity Estimation and Reconstruction of hidden components from noisy periodic signal

Periodicity estimation from an arbitrary length noisy signal is computationally very costly. A recently developed Ramanujan Fat Dictionary is one of the ways to find the hidden components from an arbitrary length (non integral multiple of period) of the signal. This method suffers from high run time due to the lack of information about the period and effect of noise on the signal. We propose a new method that efficiently estimates the period of the signal and finding the hidden components thus becomes easy from it. Our method works well with significantly low SNR values and runs in O(n) time complexity, n being length of the signal. Comparision of run time analysis between our method for period estimation of a given signal and SVD method at various SNR values has been made and the corresponding hidden components are there by extracted by projecting onto the factor-Ramanujan Subspaces.

eess.SP↗

arXiv2016

An Analysis of Rhythmic Staccato-Vocalization Based on Frequency Demodulation for Laughter Detection in Conversational Meetings

Human laugh is able to convey various kinds of meanings in human communications. There exists various kinds of human laugh signal, for example: vocalized laugh and non vocalized laugh. Following the theories of psychology, among all the vocalized laugh type, rhythmic staccato-vocalization significantly evokes the positive responses in the interactions. In this paper we attempt to exploit this observation to detect human laugh occurrences, i.e., the laughter, in multiparty conversations from the AMI meeting corpus. First, we separate the high energy frames from speech, leaving out the low energy frames through power spectral density estimation. We borrow the algorithm of rhythm detection from the area of music analysis to use that on the high energy frames. Finally, we detect rhythmic laugh frames, analyzing the candidate rhythmic frames using statistics. This novel approach for detection of `positive' rhythmic human laughter performs better than the standard laughter classification baseline.

cs.SD↗