arXiv subjects

Irfan Khan

Publications and source records attributed to Irfan Khan.

14 recordsLinked to original sources

arXiv2026

EMMI: Edge Multi-Modal Intelligence for Communication-Efficient MLLM Inference via Fused Representation Compression

Recent advances in multimodal large language mod- els (MLLMs) have opened new opportunities for edge intelligence by enabling reasoning across heterogeneous sensor modalities, such as vision, text, and telemetry data. However, deploying these capabilities on resource-constrained edge platforms remains challenging due to the substantial computational, memory, and communication demands of modern MLLMs. Rather than transmitting raw sensor observations or partitioning neural networks at intermediate layers, Edge Multi-Modal Intelligence (EMMI) communicates a compact representation between edge devices and server resources, enabling communication-efficient edge MLLM inference. To achieve this, EMMI performs modality-specific encoding, cross-modal representation fusion, and learned compression at the edge, transmitting only a compact latent representation to server-side resources for high-capacity MLLM reasoning. This representation-centric design reduces communication overhead, preserves local data privacy, and provides a fixed-size interface between heterogeneous edge devices and server-side MLLMs. Evaluation on a representative multimodal benchmark demonstrates that EMMI can reduce the communication payload by 32x while maintaining comparable downstream accuracy, resulting in up to a 3.4x reduction in estimated end-to-end inference latency under bandwidth-constrained edge conditions.

cs.LG↗

arXiv2025

Prmpt2Adpt: Prompt-Based Zero-Shot Domain Adaptation for Resource-Constrained Environments

Unsupervised Domain Adaptation (UDA) is a critical challenge in real-world vision systems, especially in resource-constrained environments like drones, where memory and computation are limited. Existing prompt-driven UDA methods typically rely on large vision-language models and require full access to source-domain data during adaptation, limiting their applicability. In this work, we propose Prmpt2Adpt, a lightweight and efficient zero-shot domain adaptation framework built around a teacher-student paradigm guided by prompt-based feature alignment. At the core of our method is a distilled and fine-tuned CLIP model, used as the frozen backbone of a Faster R-CNN teacher. A small set of low-level source features is aligned to the target domain semantics-specified only through a natural language prompt-via Prompt-driven Instance Normalization (PIN). These semantically steered features are used to briefly fine-tune the detection head of the teacher model. The adapted teacher then generates high-quality pseudo-labels, which guide the on-the-fly adaptation of a compact student model. Experiments on the MDS-A dataset demonstrate that Prmpt2Adpt achieves competitive detection performance compared to state-of-the-art methods, while delivering up to 7x faster adaptation and 5x faster inference speed using few source images-making it a practical and scalable solution for real-time adaptation in low-resource domains.

cs.CV↗

arXiv2025

XG-NID: Dual-Modality Network Intrusion Detection using a Heterogeneous Graph Neural Network and Large Language Model

In the rapidly evolving field of cybersecurity, the integration of flow-level and packet-level information for real-time intrusion detection remains a largely untapped area of research. This paper introduces "XG-NID," a novel framework that, to the best of our knowledge, is the first to fuse flow-level and packet-level data within a heterogeneous graph structure, offering a comprehensive analysis of network traffic. Leveraging a heterogeneous graph neural network (GNN) with graph-level classification, XG-NID uniquely enables real-time inference while effectively capturing the intricate relationships between flow and packet payload data. Unlike traditional GNN-based methodologies that predominantly analyze historical data, XG-NID is designed to accommodate the heterogeneous nature of network traffic, providing a robust and real-time defense mechanism. Our framework extends beyond mere classification; it integrates Large Language Models (LLMs) to generate detailed, human-readable explanations and suggest potential remedial actions, ensuring that the insights produced are both actionable and comprehensible. Additionally, we introduce a new set of flow features based on temporal information, further enhancing the contextual and explainable inferences provided by our model. To facilitate practical application and accessibility, we developed "GNN4ID," an open-source tool that enables the extraction and transformation of raw network traffic into the proposed heterogeneous graph structure, seamlessly integrating flow and packet-level data. Our comprehensive quantitative comparative analysis demonstrates that XG-NID achieves an F1 score of 97\% in multi-class classification, outperforming existing baseline and state-of-the-art methods. This sets a new standard in Network Intrusion Detection Systems by combining innovative data fusion with enhanced interpretability and real-time capabilities.

cs.CR↗

arXiv2024

A Composite Hydrogel of Porous Gold Nanorods and Gelatin: Nanoscale Structure and Rheo-Mechanical Properties

Incorporating nanomaterials into hydrogels allows for the creation of versatile materials with properties that can be precisely tailored by manipulating their nanoscale structures, leading to a wide range of bulk properties. Investigating the structural and property characteristics of composite hydrogels is crucial in tailoring their performance for specific applications. This study focuses on investigating the correlation between the structural arrangement and properties of a composite hydrogel of thermoresponsive polymer, gelatin, and light-responsive antimicrobial porous gold nanorods, $PAuNR$. The rheo-mechanical properties of the composite hydrogels are correlated with their nanoscale structural characteristics, investigated using small-angle neutron scattering ($SANS$). Analysis of $SANS$ data reveals a decrease in the fractal dimension of $PAuNRs$ incorporated hydrogel matrix, as compared to pure gelatin. Incorporating $PAuNRs$ results in formation of softer composite hydrogel as evident from decrease in viscoelastic moduli, critical yield strain, denaturation temperature and swelling ratio. Our results demonstrates that the structural modulation at the nanoscale can be precisely controlled through adjusting $PAuNRs$ concentration and temperature providing an fabrication mechanism for hydrogels with desired elastic properties. The reduced elasticity of the composite hydrogel and light sensitive/antimicrobial property of the $PAuNRs$ makes this system suitable for specific biomedical applications, such as tissue engineering, device fabrication and stimuli based controlled drug delivery devices respectively.

cond-mat.soft↗

arXiv2024

ByteStack-ID: Integrated Stacked Model Leveraging Payload Byte Frequency for Grayscale Image-based Network Intrusion Detection

In the ever-evolving realm of network security, the swift and accurate identification of diverse attack classes within network traffic is of paramount importance. This paper introduces "ByteStack-ID," a pioneering approach tailored for packet-level intrusion detection. At its core, ByteStack-ID leverages grayscale images generated from the frequency distributions of payload data, a groundbreaking technique that greatly enhances the model's ability to discern intricate data patterns. Notably, our approach is exclusively grounded in packet-level information, a departure from conventional Network Intrusion Detection Systems (NIDS) that predominantly rely on flow-based data. While building upon the fundamental concept of stacking methodology, ByteStack-ID diverges from traditional stacking approaches. It seamlessly integrates additional meta learner layers into the concatenated base learners, creating a highly optimized, unified model. Empirical results unequivocally confirm the outstanding effectiveness of the ByteStack-ID framework, consistently outperforming baseline models and state-of-the-art approaches across pivotal performance metrics, including precision, recall, and F1-score. Impressively, our proposed approach achieves an exceptional 81\% macro F1-score in multiclass classification tasks. In a landscape marked by the continuous evolution of network threats, ByteStack-ID emerges as a robust and versatile security solution, relying solely on packet-level information extracted from network traffic data.

cs.CR↗

arXiv2023

Detecting Unknown Attacks in IoT Environments: An Open Set Classifier for Enhanced Network Intrusion Detection

The widespread integration of Internet of Things (IoT) devices across all facets of life has ushered in an era of interconnectedness, creating new avenues for cybersecurity challenges and underscoring the need for robust intrusion detection systems. However, traditional security systems are designed with a closed-world perspective and often face challenges in dealing with the ever-evolving threat landscape, where new and unfamiliar attacks are constantly emerging. In this paper, we introduce a framework aimed at mitigating the open set recognition (OSR) problem in the realm of Network Intrusion Detection Systems (NIDS) tailored for IoT environments. Our framework capitalizes on image-based representations of packet-level data, extracting spatial and temporal patterns from network traffic. Additionally, we integrate stacking and sub-clustering techniques, enabling the identification of unknown attacks by effectively modeling the complex and diverse nature of benign behavior. The empirical results prominently underscore the framework's efficacy, boasting an impressive 88\% detection rate for previously unseen attacks when compared against existing approaches and recent advancements. Future work will perform extensive experimentation across various openness levels and attack scenarios, further strengthening the adaptability and performance of our proposed solution in safeguarding IoT environments.

cs.CR↗

arXiv2023

Statistical LOS/NLOS Classification for UWB Channels

Ultrawideband (UWB) technology has attracted a lot of attention for indoor and outdoor positioning systems due to its high accuracy and robustness in non-line-of-sight (NLOS) environments. However, UWB signals are affected by multipath propagation which causes errors in localization. To overcome this problem, researchers have proposed various techniques for NLOS identification and mitigation. One of the approaches is statistical LOS/NLOS classification, which uses statistical parameters of the received signal to distinguish between LOS and NLOS channels. In this paper, we formulated several techniques which can be used for effectively classifying a Line of Sight (LOS) channel from a Non-Line of Sight (NLOS) channel. Various parameters obtained from Channel Impulse Response (CIR) like Skewness, Kurtosis, Root Mean Squared Delay Spread (RDS), Mean Excess Delay (MED), Energy, Energy Ratio, and Mean of Covariance Matrix are used for channel classification. In addition to this, the Joint Probability Density Functions (PDFs) of various parameters are used to improve the accuracy of UWB LOS/NLOS channel classification. Two different criteria-Likelihood Ratio and Hypothesis Tests are used for the identification of the channel.

eess.SP↗

arXiv2022

Quantum hardware calculations of periodic systems with partition-measurement symmetry verification: simplified models of hydrogen chain and iron crystals

Running quantum algorithms on real hardware is essential for understanding their strengths and limitations, especially in the noisy intermediate scale quantum (NISQ) era. Herein we focus on the practical aspect of quantum computational calculations of solid-state crystalline materials based on theory developed in our group by using real quantum hardware with a novel noise mitigation technique referred to as partition-measurement symmetry verification, which performs post-selection of shot counts based on $Z_{2}$ and $U_{1}$ symmetry verification. We select two periodic systems with different level of complexity for these calculations. One of them is the distorted hydrogen chain as an example of very simple systems, and the other one is iron crystal in the BCC and FCC phases as it is considered to be inaccessible by using classical computational wavefunction methods. The ground state energies are evaluated based on the translational quantum subspace expansion (TransQSE) method for the hydrogen chain, and periodic boundary condition adapted VQE for our iron models. By applying these techniques for the simplest 2 qubit iron model systems, the correlation energies obtained by the hardware calculations agree with those of the state-vector simulations within $\sim$5 kJ/mol. Although the quantum computational resources used for those experiments are still limited, the techniques applied to obtain our simplified models will be applicable in essentially the same manner to more complicated cases as quantum hardware matures.

quant-ph↗

arXiv2021

Explainable Signature-based Machine Learning Approach for Identification of Faults in Grid-Connected Photovoltaic Systems

The transformation of conventional power networks into smart grids with the heavy penetration level of renewable energy resources, particularly grid-connected Photovoltaic (PV) systems, has increased the need for efficient fault identification systems. Malfunctioning any single component in grid-connected PV systems may lead to grid instability and other serious consequences, showing that a reliable fault identification system is the utmost requirement for ensuring operational integrity. Therefore, this paper presents a novel fault identification approach based on statistical signatures of PV operational states. These signatures are unique because each fault has a different nature and distinctive impact on the electrical system. Thus, the Random Forest Classifier trained on these extracted signatures showed 100% accuracy in identifying all types of faults. Furthermore, the performance comparison of the proposed framework with other Machine Learning classifiers depicts its credibility. Moreover, to elevate user trust in the predicted outcomes, SHAP (Shapley Additive Explanation) was utilized during the training phase to extract a complete model response (global explanation). This extracted global explanation can help in the assessment of predicted outcomes credibility by decoding each prediction in terms of features contribution. Hence, the proposed explainable signature-based fault identification technique is highly credible and fulfills all the requirements of smart grids.

cs.LG↗

arXiv2021

An Autonomous Self-Incremental Learning Approach for Detection of Cyber Attacks on Unmanned Aerial Vehicles (UAVs)

As the technological advancement and capabilities of automated systems have increased drastically, the usage of unmanned aerial vehicles for performing human-dependent tasks without human indulgence has also spiked. Since unmanned aerial vehicles are heavily dependent on Information and Communication Technology, they are highly prone to cyber-attacks. With time more advanced and new attacks are being developed and employed. However, the current Intrusion detection system lacks detection and classification of new and unknown attacks. Therefore, for having an autonomous and reliable operation of unmanned aerial vehicles, more robust and automated cyber detection and protection schemes are needed. To address this, we have proposed an autonomous self-incremental learning architecture, capable of detecting known and unknown cyber-attacks on its own without any human interference. In our approach, we have combined signature-based detection along with anomaly detection in such a way that the signature-based detector autonomously updates its attack classes with the help of an anomaly detector. To achieve this, we have implemented an incremental learning approach, updating our model to incorporate new classes without forgetting the old ones. To validate the applicability and effectiveness of our proposed architecture, we have implemented it in a trial scenario and then compared it with the traditional offline learning approach. Moreover, our anomaly-based detector has achieved a 100% detection rate for attacks.

cs.CR↗

arXiv2021

A Sequential Supervised Machine Learning Approach for Cyber Attack Detection in a Smart Grid System

Modern smart grid systems are heavily dependent on Information and Communication Technology, and this dependency makes them prone to cyberattacks. The occurrence of a cyberattack has increased in recent years resulting in substantial damage to power systems. For a reliable and stable operation, cyber protection, control, and detection techniques are becoming essential. Automated detection of cyberattacks with high accuracy is a challenge. To address this, we propose a two-layer hierarchical machine learning model having an accuracy of 95.44 % to improve the detection of cyberattacks. The first layer of the model is used to distinguish between the two modes of operation (normal state or cyberattack). The second layer is used to classify the state into different types of cyberattacks. The layered approach provides an opportunity for the model to focus its training on the targeted task of the layer, resulting in improvement in model accuracy. To validate the effectiveness of the proposed model, we compared its performance against other recent cyber attack detection models proposed in the literature.

cs.CR↗

arXiv2018

Compressive Sensing and Morphology Singular Entropy-Based Real-time Secondary Voltage Control of Multi-area Power Systems

This paper presents an improved secondary voltage control (SVC) methodology incorporating compressive sensing (CS) for a multi-area power system. SVC minimizes the voltage deviation of the load buses while CS deals with the problem of the limited bandwidth capacity of the communication channel by reducing the size of massive data output from phasor measurement unit (PMU) based monitoring system. The proposed strategy further incorporates the application of a Morphological Median Filter (MMF) to reduce noise from the output of the PMUs. To keep the control area secure and protected locally, Mathematical Singular Entropy (MSE) based fault identification approach is utilized for fast discovery of faults in the control area. Simulation results with 27-bus and 486-bus power systems show that CS can reduce the data size up to 1/10th while the MSE based fault identification technique can accurately distinguish between fault and steady state conditions.

eess.SP↗

arXiv2018

Effect of the Approximation of Voltage Angle Difference on the OPF algorithms in the Power Network

In real-time applications involving power flow equations, measuring of voltage phase angle difference of the connected buses is essential. However, it needs special techniques to measure voltage angle difference, which may enlarge the computational burden of the working controller and hence, may make the control process slow. In this paper, authors investigate the approximation of angle difference to zero and its effects on the convergence speed and optimal solutions of a distributed algorithm. To test this approximation, a distributed nonlinear algorithm is proposed to optimize the multi-objective function which includes power loss, voltage deviation and cost of reactive power generation, by controlling the reactive power generations from distributed generators. Authors investigate the reasons which may outlaw making this approximation and finally, propose a condition to make such approximation. Importance of making this approximation in terms of fast convergence of the algorithms is also illustrated.

eess.SP↗

arXiv2018

A non-linear convex cost model for economic dispatch in microgrids

This paper proposes a convex non-linear cost saving model for optimal economic dispatch in a microgrid. The mod-el incorporates energy storage degradation cost and intermittent renewable generation. Cell degradation cost being a non-linear model, its incorporation in an objective function alters the convexity of the optimization problem and stochastic algorithms are required for its solution. This paper builds on the scope for usage of macroscopically semi-empirical models for degradation cost in economic dispatch problems and proves that these cost models derived from the existing semi-empirical capacity fade equations for LiFePO4 cells are convex under some operating condi-tions. The proposed non-linear model was tested on two data sets of varying size which portray different trends of seasonality. The results show that the model reflects the trends of seasonality existing in the data sets and it mini-mizes the total fuel cost globally when compared to conventional systems of economic dispatch. The results thus indicate that the model achieves a more accurate estimate of fuel cost in the system and can be effectively utilized for cost analysis in power system applications.

eess.SP↗