arXiv subjects

Debayan Das

Publications and source records attributed to Debayan Das.

At least 19 recordsLinked to original sources

arXiv2025

TYTAN: Taylor-series based Non-Linear Activation Engine for Deep Learning Accelerators

The rapid advancement in AI architectures and the proliferation of AI-enabled systems have intensified the need for domain-specific architectures that enhance both the acceleration and energy efficiency of AI inference, particularly at the edge. This need arises from the significant resource constraints-such as computational cost and energy consumption-associated with deploying AI algorithms, which involve intensive mathematical operations across multiple layers. High-power-consuming operations, including General Matrix Multiplications (GEMMs) and activation functions, can be optimized to address these challenges. Optimization strategies for AI at the edge include algorithmic approaches like quantization and pruning, as well as hardware methodologies such as domain-specific accelerators. This paper proposes TYTAN: TaYlor-series based non-linear acTivAtion eNgine, which explores the development of a Generalized Non-linear Approximation Engine (G-NAE). TYTAN targets the acceleration of non-linear activation functions while minimizing power consumption. The TYTAN integrates a re-configurable hardware design with a specialized algorithm that dynamically estimates the necessary approximation for each activation function, aimed at achieving minimal deviation from baseline accuracy. The proposed system is validated through performance evaluations with state-of-the-art AI architectures, including Convolutional Neural Networks (CNNs) and Transformers. Results from system-level simulations using Silvaco's FreePDK45 process node demonstrate TYTAN's capability to operate at a clock frequency >950 MHz, showcasing its effectiveness in supporting accelerated, energy-efficient AI inference at the edge, which is ~2 times performance improvement, with ~56% power reduction and ~35 times lower area compared to the baseline open-source NVIDIA Deep Learning Accelerator (NVDLA) implementation.

cs.AR↗

arXiv2025

Machine Learning Power Side-Channel Attack on SNOW-V

This paper demonstrates a power analysis-based Side-Channel Analysis (SCA) attack on the SNOW-V encryption algorithm, which is a 5G mobile communication security standard candidate. Implemented on an STM32 microcontroller, power traces captured with a ChipWhisperer board were analyzed, with Test Vector Leakage Assessment (TVLA) confirming exploitable leakage. Profiling attacks using Linear Discriminant Analysis (LDA) and Fully Connected Neural Networks (FCN) achieved efficient key recovery, with FCN achieving > 5X lower minimum traces to disclosure (MTD) compared to the state-of-the-art Correlational Power Analysis (CPA) assisted with LDA. The results highlight the vulnerability of SNOW-V to machine learning-based SCA and the need for robust countermeasures.

cs.CR↗

arXiv2025

From the Two-Capacitor Paradox to Electromagnetic Side-Channel Mitigation in Digital Circuits

The classical two-capacitor paradox of the lost energy is revisited from an electronic circuit security stand-point. The paradox has been solved previously by various researchers, and the energy lost during the charging of capacitors has been primarily attributed to the heat and radiation. We analytically prove this for various standard resistor-capacitor (RC) and resistor-inductor-capacitor (RLC) circuit models. From the perspective of electronic system security, electromagnetic (EM) side-channel analysis (SCA) has recently gained significant prominence with the growth of resource-constrained, internet connected devices. This article connects the energy lost due to capacitor charging to the EM SCA leakage in electronic devices, leading to the recovery of the secret encryption key embedded within the device. Finally, with an understanding of how lost energy relates to EM radiation, we propose adiabatic charging as a solution to minimize EM leakage, thereby paving the way towards low-overhead EM SCA resilience.

cs.CR↗

arXiv2025

SlimEdge: Performance and Device Aware Distributed DNN Deployment on Resource-Constrained Edge Hardware

Distributed deep neural networks (DNNs) have become central to modern computer vision, yet their deployment on resource-constrained edge devices remains hindered by substantial parameter counts, computational demands, and the probability of device failure. Here, we present an approach to the efficient deployment of distributed DNNs that jointly respect hardware limitations, preserve task performance, and remain robust to partial system failures. Our method integrates structured model pruning with a multi-objective optimization framework to tailor network capacity for heterogeneous device constraints, while explicitly accounting for device availability and failure probability during deployment. We demonstrate this framework using Multi-View Convolutional Neural Networks (MVCNN), a state-of-the-art architecture for 3D object recognition, by quantifying the contribution of individual views to classification accuracy and allocating pruning budgets accordingly. Experimental results show that the resulting models satisfy user-specified bounds on accuracy and memory footprint, even under multiple simultaneous device failures. The inference time is reduced by factors up to 4.7x across diverse simulated device configurations. These findings suggest that performance-aware, view-adaptive, and failure-resilient compression provides a viable pathway for deploying complex vision models in distributed edge environments.

cs.DC↗

arXiv2025

A Fast, Parallelized, GPU-Accelerated Photochemical Model, XODIAC, with Built-in Equilibrium Chemistry and Multiple Chemical Networks for Exoplanetary Atmospheres

The launch of the James Webb Space Telescope (JWST) has delivered high-quality atmospheric observations and expanded the known chemical inventory of exoplanetary atmospheres, opening new avenues for atmospheric chemistry modeling to interpret these data. Here, we present XODIAC, a fast, GPU-accelerated, one-dimensional photochemical model with a built-in equilibrium chemistry solver, an updated thermochemical database, and three chemical reaction networks. This framework enables comparative atmospheric chemistry studies, including the newly developed XODIAC-2025 network, a state-of-the-art C-H-O-N-P-S-Metals network, linking 594 species through 7,720 reactions. The other two are existing, publicly available C-H-O-N-S and C-H-O-N-S-Metals networks, from the established photochemical models VULCAN and ARGO, respectively, which are commonly used in the community. The XODIAC model has been rigorously benchmarked on the well-studied hot Jupiter HD 189733 b, with results compared against these two models. Benchmarking shows excellent agreement and demonstrates that, when the same chemical network and initial conditions are used, the numerical scheme for solving atmospheric chemistry does not significantly affect the results. We also revisited the atmospheric chemistry of HD 189733 b and performed a comparative analysis across the three networks. Sulfur chemistry shows the least variation across networks, carbon chemistry shows slightly more, and phosphorus chemistry varies the most, primarily due to the introduction of unique PHO and PN pathways comprising 390 reactions in the XODIAC-2025 network. These findings highlight XODIAC's capability to advance exoplanetary atmospheric chemistry and provide a robust framework for comparative exoplanetology.

astro-ph.EP↗

arXiv2025

Millimetre-Wave Comb Generated by an Optical Microcomb

Metrological-grade millimetre wave baseband comb sources covering the subterahertz window are a key building block for next-generation wireless communications, precision sensing, and positioning systems. While optical microcombs have set new benchmarks in ultra-low phase noise single-frequency microwave generation, to date, no microcomb source has directly produced a millimetre-wave baseband comb. Here, we present a 50 GHz repetition rate carrier-envelope offset estabilised millimetre-wave baseband comb source covering the sub-terahertz region, generated from an optical microcomb source. Our microresonator-filtered microcomb enables direct, coherent downconversion via photoconductive antennas, even without external amplification. The metrological-grade optical soliton source produces single-cycle, naturally zero carrier-envelope offset millimetrewave baseband combs. It supports time-domain spectroscopy without any need to temporally align the source and detection pulses, as the ultra-high phase coherence allows significant differences between the optical paths of the source and detection pulses, which we tested over 8m, finding no degradation even in freerunning operation. Finally, the multisoliton operation regime provides a simple way of spectrally tailoring the microwave output by selecting different optical soliton states.

physics.optics↗

arXiv2025

A Next-Generation Exoplanet Atmospheric Retrieval Framework NEXOTRANS for Emission Spectroscopy: New Constraints and Atmospheric Characterization of WASP-69b Using JWST NIRCam and MIRI Observations

Thermal emission spectra provide key insights into the atmospheric composition and especially the temperature structure of an exoplanet. With broader wavelength coverage, sensitivity and higher resolution, JWST has enabled robust constraints on these properties, including detections of photochemical products. This advances the need for retrieval frameworks capable of navigating complex parameter spaces for accurate data interpretation. In this work, we introduce the emission retrieval module of NEXOTRANS, which employs both one- and two-stream radiative transfer approximations and leverages Bayesian and machine learning techniques for retrievals. It also incorporates approximate disequilibrium chemistry models to infer photochemical species like SO2. We applied NEXOTRANS to the JWST NIRCam and MIRI emission observations of WASP-69b, covering the 2-12 microns range. The retrievals place robust constraints on the volume mixing ratios (VMR) of H2O, CO2, CO, CH4, and potential SO2. The best-fit model, i.e, free chemistry combined with non-uniform aerosol coverage, yields a log(VMR) = -3.78 (+0.15/-0.17) for H2O and -5.77 (+0.09/-0.10) for CO2 which has a sharp absorption at 4.3 micron. The second best-fit model, the hybrid equilibrium chemistry (utilizing equilibrium chemistry-grids) combined with non-uniform aerosol yields a C/O of 0.42 (+0.17/-0.13) and a metallicity of log[M/H] = 1.24 (+0.17/-0.14), corresponding to approximately 17.38 times the solar value. This hybrid chemistry retrieval also constrain SO2 with a log(VMR) = -4.85 (+0.28/-0.29), indicating possible absorption features in the 7-8 microns range. These results highlight NEXOTRANS's capability to significantly advance JWST emission spectra interpretation, offering broader insights into exoplanetary atmospheres.

astro-ph.EP↗

arXiv2025

Topological Quenching of Noise in a Free-Running Moebius Microcomb

Microcombs require ultralow-noise repetition rates to enable next-generation applications in metrology, high-speed communications, microwave photonics, and sensing, where spectral purity is a central performance metric. Best-performing sources operate actively locked at "quiet points" in parameter space, fixed by device and material properties. Creating broad, low-noise operating regions with relaxed constraints-especially in simplified free-running architectures that avoid electronics-heavy control-remains an open challenge. Here, we demonstrate a symmetry-protected topological M\"obius soliton molecule that enables intrinsically low phase noise in a fully free-running microcomb, operating without any external referencing or control. Using a microresonator-filtered laser, we implement a M\"obius geometry via interleaved microcavity modes. Upon the formation of a topological M\"obius soliton molecule, the free-running laser exhibits over 15 dB of phase-noise suppression across 10 Hz-10 kHz at a 100 GHz repetition rate, yielding -63 dBc/Hz phase noise at 1 kHz and an Allan deviation of 4x10^-10 at 10 s average time-without any external control. We show that the M\"obius structure brings dynamic robustness to the comb, and we demonstrate a symmetry-protected topological regime that enables long-term drift-invariant operation. Our results establish a route to intrinsically noise-quenched microcombs operating in a fully free-running configuration, governed by internal physical principles and suitable for field-deployable, low-noise photonic systems.

physics.optics↗

arXiv2025

A Next-Generation Exoplanet Atmospheric Retrieval Framework for Transmission Spectroscopy (NEXOTRANS): Comparative Characterization for WASP-39 b Using JWST NIRISS, NIRSpec PRISM, and MIRI Observations

The advent of JWST has marked a new era in exoplanetary atmospheric studies, offering higher-resolution data and greater precision across a broader spectral range than previous space-based telescopes. Accurate analysis of these datasets requires advanced retrieval frameworks capable of navigating complex parameter spaces. We present NEXOTRANS, an atmospheric retrieval framework that integrates Bayesian inference using UltraNest/PyMultiNest with four machine learning algorithms: Random Forest, Gradient Boosting, K-Nearest Neighbor, and Stacking Regressor. This hybrid approach enables a comparison between traditional Bayesian methods and computationally efficient machine learning techniques. Additionally, NEXOTRANS incorporates NEXOCHEM, a module for solving equilibrium chemistry. We applied NEXOTRANS to JWST observations of the Saturn-mass exoplanet WASP-39 b, spanning wavelengths from 0.6 microns to 12.0 microns using NIRISS, NIRSpec PRISM, and MIRI. Four chemistry models - free, equilibrium, modified hybrid equilibrium, and modified equilibrium-offset chemistry - were explored to retrieve precise Volume Mixing Ratios (VMRs) for H2O, CO2, CO, H2S, and SO2. Absorption features in both NIRSpec PRISM and MIRI data constrained SO2 log VMRs to values between -6.25 and -5.73 for all models except equilibrium chemistry. High-altitude aerosols, including ZnS and MgSiO3, were inferred, with constraints on their VMRs, particle sizes, and terminator coverage fractions, providing insights into cloud composition. For the best-fit modified hybrid equilibrium model, we derived super-solar elemental abundances of O/H = 14.12 (+2.86/-1.82) x solar, C/H = 21.37 (+4.93/-3.18) x solar, and S/H = 5.37 (+0.79/-0.65) x solar, along with a C/O ratio of 1.35 (+0.05/-0.02) x solar, demonstrating NEXOTRANS's potential for atmospheric characterization in the JWST era and beyond.

astro-ph.EP↗

arXiv2024

A Voice-based Triage for Type 2 Diabetes using a Conversational Virtual Assistant in the Home Environment

Incorporating cloud technology with Internet of Medical Things for ubiquitous healthcare has seen many successful applications in the last decade with the advent of machine learning and deep learning techniques. One of these applications, namely voice-based pathology, has yet to receive notable attention from academia and industry. Applying voice analysis to early detection of fatal diseases holds much promise to improve health outcomes and quality of life of patients. In this paper, we propose a novel application of acoustic machine learning based triaging into commoditised conversational virtual assistant systems to pre-screen for onset of diabetes. Specifically, we developed a triaging system which extracts acoustic features from the voices of n=24 older adults when they converse with a virtual assistant and predict the incidence of Diabetes Mellitus (Type 2) or not. Our triaging system achieved hit-rates of 70% and 60% for male and female older adult subjects, respectively. Our proposed triaging uses 7 non-identifiable voice-based features and can operate within resource-constrained embedded systems running voice-based virtual assistants. This application demonstrates the feasibility of applying voice-based pathology analysis to improve health outcomes of older adults within the home environment by early detection of life-changing chronic conditions like diabetes.

cs.SD↗

arXiv2024

R-STELLAR: A Resilient Synthesizable Signature Attenuation SCA Protection on AES-256 with built-in Attack-on-Countermeasure Detection

Side channel attacks (SCAs) remain a significant threat to the security of cryptographic systems in modern embedded devices. Even mathematically secure cryptographic algorithms, when implemented in hardware, inadvertently leak information through physical side channel signatures such as power consumption, electromagnetic (EM) radiation, light emissions, and acoustic emanations. Exploiting these side channels significantly reduces the search space of the attacker. In recent years, physical countermeasures have significantly increased the minimum traces to disclosure (MTD) to 1 billion. Among them, signature attenuation is the first method to achieve this mark. Signature attenuation often relies on analog techniques, and digital signature attenuation reduces MTD to 20 million, requiring additional methods for high resilience. We focus on improving the digital signature attenuation by an order of magnitude (MTD 200M). Additionally, we explore possible attacks against signature attenuation countermeasure. We introduce a Voltage drop Linear region Biasing (VLB) attack technique that reduces the MTD to over 2000 times less than the previous threshold. This is the first known attack against a physical side-channel attack (SCA) countermeasure. We have implemented an attack detector with a response time of 0.8 milliseconds to detect such attacks, limiting SCA leakage window to sub-ms, which is insufficient for a successful attack.

cs.CR↗

arXiv2024

SNOW-SCA: ML-assisted Side-Channel Attack on SNOW-V

This paper presents SNOW-SCA, the first power side-channel analysis (SCA) attack of a 5G mobile communication security standard candidate, SNOW-V, running on a 32-bit ARM Cortex-M4 microcontroller. First, we perform a generic known-key correlation (KKC) analysis to identify the leakage points. Next, a correlation power analysis (CPA) attack is performed, which reduces the attack complexity to two key guesses for each key byte. The correct secret key is then uniquely identified utilizing linear discriminant analysis (LDA). The profiled SCA attack with LDA achieves 100% accuracy after training with $<200$ traces, which means the attack succeeds with just a single trace. Overall, using the \textit{combined CPA and LDA attack} model, the correct secret key byte is recovered with <50 traces collected using the ChipWhisperer platform. The entire 256-bit secret key of SNOW-V can be recovered incrementally using the proposed SCA attack. Finally, we suggest low-overhead countermeasures that can be used to prevent these SCA attacks.

cs.CR↗

arXiv2023

A 334$\mu$W 0.158mm$^2$ ASIC for Post-Quantum Key-Encapsulation Mechanism Saber with Low-latency Striding Toom-Cook Multiplication Authors Version

The hard mathematical problems that assure the security of our current public-key cryptography (RSA, ECC) are broken if and when a quantum computer appears rendering them ineffective for use in the quantum era. Lattice based cryptography is a novel approach to public key cryptography, of which the mathematical investigation (so far) resists attacks from quantum computers. By choosing a module learning with errors (MLWE) algorithm as the next standard, National Institute of Standard & Technology (NIST) follows this approach. The multiplication of polynomials is the central bottleneck in the computation of lattice based cryptography. Because public key cryptography is mostly used to establish common secret keys, focus is on compact area, power and energy budget and to a lesser extent on throughput or latency. While most other work focuses on optimizing number theoretic transform (NTT) based multiplications, in this paper we highly optimize a Toom-Cook based multiplier. We demonstrate that a memory-efficient striding Toom-Cook with lazy interpolation, results in a highly compact, low power implementation, which on top enables a very regular memory access scheme. To demonstrate the efficiency, we integrate this multiplier into a Saber post-quantum accelerator, one of the four NIST finalists. Algorithmic innovation to reduce active memory, timely clock gating and shift-add multiplier has helped to achieve 38% less power than state-of-the art PQC core, 4x less memory, 36.8% reduction in multiplier energy and 118x reduction in active power with respect to state-of-the-art Saber accelerator (not silicon verified). This accelerator consumes 0.158mm2 active area which is lowest reported till date despite process disadvantages of the state-of-the-art designs.

cs.CR↗

arXiv2022

A Gait Triaging Toolkit for Overlapping Acoustic Events in Indoor Environments

Gait has been used in clinical and healthcare applications to assess the physical and cognitive health of older adults. Acoustic based gait detection is a promising approach to collect gait data of older adults passively and non-intrusively. However, there has been limited work in developing acoustic based gait detectors that can operate in noisy polyphonic acoustic scenes of homes and care homes. We attribute this to the lack of good quality gait datasets from the real-world to train a gait detector on. In this paper, we put forward a novel machine learning based filter which can triage gait audio samples suitable for training machine learning models for gait detection. The filter achieves this by eliminating noisy samples at an f(1) score of 0.85 and prioritising gait samples with distinct spectral features and minimal noise. To demonstrate the effectiveness of the filter, we train and evaluate a deep learning model on gait datasets collected from older adults with and without applying the filter. The model registers an increase of 25 points in its f(1) score on unseen real-word gait data when trained with the filtered gait samples. The proposed filter will help automate the task of manual annotation of gait samples for training acoustic based gait detection models for older adults in indoor environments.

cs.SD↗

arXiv2022

Exploring extreme thermodynamics in nanoliter volumes through stimulated Brillouin-Mandelstam scattering

Examining the physical properties of materials - particularly of toxic liquids - under a wide range of thermodynamic states is a challenging problem due to the extreme conditions the material has to be exposed to. Such temperature and pressure regimes, which result in a change of refractive index and sound velocity can be accessed by optoacoustic interactions such as Brillouin-Mandelstam scattering. Here we experimentally demonstrate Brillouin-Mandelstam measurements of nanoliter volumes of liquids in extreme thermodynamic regimes. We use a fully-sealed liquid-core optical fiber containing carbon disulfide; within this waveguide, which exhibits tight optoacoustic confinement and a high Brillouin gain of (32.2 $\pm$ 0.8) 1/(Wm), we are able to conduct spatially resolved measurements of the Brillouin frequency shift. Knowledge of the local Brillouin response enables us to control the temperature and pressure independently over a wide range. We observe and measure the material properties of the liquid core at very large positive pressures (above 1000 bar), substantial negative pressures (below -300 bar) and we explore the isobaric and isochoric regimes. The extensive thermodynamic control allows the tunability of the Brillouin frequency shift of more than 40% using only minute volumes of liquid. This work opens the way for future studies of liquids under a variety of conventionally hard-to-reach conditions.

physics.optics↗

arXiv2022

A 333.9uW 0.158mm$^2$ Saber Learning with Rounding based Post-Quantum Crypto Accelerator

National Institute of Standard & Technology (NIST) is currently running a multi-year-long standardization procedure to select quantum-safe or post-quantum cryptographic schemes to be used in the future. Saber is the only LWR based algorithm to be in the final of Round 3. This work presents a Saber ASIC which provides 1.37X power-efficient, 1.75x lower area, and 4x less memory implementation w.r.t. other SoA PQC ASIC. The energy-hungry multiplier block is 1.5x energyefficient than SoA.

cs.CR↗

arXiv2020

EM-X-DL: Efficient Cross-Device Deep Learning Side-Channel Attack with Noisy EM Signatures

This work presents a Cross-device Deep-Learning based Electromagnetic (EM-X-DL) side-channel analysis (SCA), achieving >90% single-trace attack accuracy on AES-128, even in the presence of significantly lower signal-to-noise ratio (SNR), compared to the previous works. With an intelligent selection of multiple training devices and proper choice of hyperparameters, the proposed 256-class deep neural network (DNN) can be trained efficiently utilizing pre-processing techniques like PCA, LDA, and FFT on the target encryption engine running on an 8-bit Atmel microcontroller. Finally, an efficient end-to-end SCA leakage detection and attack framework using EM-X-DL demonstrates high confidence of an attacker with <20 averaged EM traces.

cs.CR↗

arXiv2020

Physical Time-Varying Transfer Functions as Generic Low-Overhead Power-SCA Countermeasure

Mathematically-secure cryptographic algorithms leak significant side channel information through their power supplies when implemented on a physical platform. These side channel leakages can be exploited by an attacker to extract the secret key of an embedded device. The existing state-of-the-art countermeasures mainly focus on the power balancing, gate-level masking, or signal-to-noise (SNR) reduction using noise injection and signature attenuation, all of which suffer either from the limitations of high power/area overheads, performance degradation or are not synthesizable. In this article, we propose a generic low-overhead digital-friendly power SCA countermeasure utilizing physical Time-Varying Transfer Functions (TVTF) by randomly shuffling distributed switched capacitors to significantly obfuscate the traces in the time domain. System-level simulation results of the TVTF-AES implemented in TSMC 65nm CMOS technology show > 4000x MTD improvement over the unprotected implementation with nearly 1.25x power and 1.2x area overheads, and without any performance degradation.

cs.CR↗