arXiv subjects

Amitabh Das

Publications and source records attributed to Amitabh Das.

15 recordsLinked to original sources

arXiv2026

Formal Verification of Secure Encrypted Virtualization

Trusted execution environments (TEEs) provide a secure environment for data and code in use, ensuring that they are protected with respect to confidentiality and integrity. Virtual machine (VM)-based TEEs utilize virtualization technology to create isolated execution spaces that can support a complete operating system or specific applications. AMD secure encrypted virtualization (SEV) is a key technology used in confidential computing in the cloud enabling hardware-based memory encryption to protect sensitive data within VMs. However, AMD SEV often operate without formal assurances of their security guarantees. Our research introduces a formal framework for representing and verifying AMD SEV confidential VMs. Specifically, we conduct design-level and property-level abstraction on AMD SEV specification and conduct property checking on the model to ensure confidentiality, integrity and availability. This approach provides a rigorous foundation for defining and verifying key security attributes for safeguarding execution environments.

cs.CR↗

arXiv2025

Exploiting Unstructured Sparsity in Fully Homomorphic Encrypted DNNs

The deployment of deep neural networks (DNNs) in privacy-sensitive environments is constrained by computational overheads in fully homomorphic encryption (FHE). This paper explores unstructured sparsity in FHE matrix multiplication schemes as a means of reducing this burden while maintaining model accuracy requirements. We demonstrate that sparsity can be exploited in arbitrary matrix multiplication, providing runtime benefits compared to a baseline naive algorithm at all sparsity levels. This is a notable departure from the plaintext domain, where there is a trade-off between sparsity and the overhead of the sparse multiplication algorithm. In addition, we propose three sparse multiplication schemes in FHE based on common plaintext sparse encodings. We demonstrate the performance gain is scheme-invariant; however, some sparse schemes vastly reduce the memory storage requirements of the encrypted matrix at high sparsity values. Our proposed sparse schemes yield an average performance gain of 2.5x at 50% unstructured sparsity, with our multi-threading scheme providing a 32.5x performance increase over the equivalent single-threaded sparse computation when utilizing 64 cores.

cs.CR↗

arXiv2023

Origin of magnetic ordering in half-Heusler RuMnGa

The half-Heusler alloy RuMnGa having valence electron count (VEC) 18 has recently been theoretically proposed to exhibit compensated ferrimagnetic (CFiM) character instead of the expected nonmagnetic ground state. On the other hand, a preliminary experimental study proposed ferromagnetic (FM) ordering. As no half-Heusler system with VEC 18 is known to exhibit magnetic ordering, we have investigated the details of crystal structure and magnetic properties of RuMnGa using a combination of experimental tools, viz., x-ray and neutron diffraction techniques, dc and ac susceptibility, isothermal magnetisation, heat capacity, resistivity and neutron depolarisation measurements. Rietveld refinements of x-ray and neutron diffraction data suggest single phase nature of the compound with elemental composition RuMn$_{0.86}$Ga$_{1.14}$. We have shown that the system exhibits FM-type ordering owing to the inherent presence of this minor off-stoichiometry, showing very low magnetic moment. The system also exhibits reentrant canonical spin-glass behaviour, which is rarely observed in half-Heusler alloys. The temperature coefficient of resistivity changes its sign from negative to positive and further to negative as the temperature decreases.

cond-mat.mtrl-sci↗

arXiv2023

Experimental observation of spin glass state in highly disordered quaternary Heusler alloy FeRuMnGa

The realization of spin-glass (S-G) state in Heusler alloys is very rare despite the presence of inherent structural and elemental disorder in those compounds. Although a few half and full Heusler alloys are known to exhibit S-G state, there is hardly any manifestation of the same in cases of quaternary Heusler compounds. Here we report the observation of S-G state in a highly disordered equiatomic quaternary Heusler compound: FeRuMnGa, where the S-G state is in between of canonical S-G and cluster glass. Different intricate features of S-G state including non-equilibrium magnetic dynamics at low temperature in the compound are unveiled through our comprehensive magnetic, heat capacity and neutron diffraction studies. The structural disorder in the sample is neither conventional \textit{A2}- nor \textit{B2}-type while those two types are commonly observed for Heusler compounds. The presence of disorder also plays a significant role in electron transport properties of the alloy, which is reflected in its exhibition of semi-metallic behavior and anomalous Hall effect at low temperature.

cond-mat.mtrl-sci↗

arXiv2023

4d-element induced improvement of structural disorder and development of weakly re-entrant spin-glass behaviour in NiRuMnSn

The pursuit of efficient spin-polarization in quaternary Heusler alloys with the general formula $XX'YZ$ (where X, $X'$, and Y are transition metals and Z is a p-block element), has been a subject of significant scientific interest. While previous studies shows that isoelectronic substitution of 4d element in place of 3d element in quaternary Heusler alloy, improves the half-metallic ferromagnetic characteristics, our research on the quaternary Heusler alloy NiRuMnSn reveals a strikingly different scenario. In this study, we present a detailed structural analysis of the material using X-ray absorption fine structure (EXAFS) and neutron diffraction (ND) techniques, which confirms the formation of a single-phase compound with 50:50 site disorder between Ni/Ru atoms at 4c/4d sites. Contrary to expectations, our DFT calculations suggests a considerable decrease in spin-polarization even in the ordered structure. Additionally, we report on the compound's exceptional behavior, displaying a rare re-entrant spin glass property below $\sim$60 K, a unique and intriguing feature for quaternary Heusler-type compounds.

cond-mat.mtrl-sci↗

arXiv2022

Extrinsic to intrinsic mechanism crossover of anomalous Hall effect in the Ir-doped MnPtSn Heusler system

Recent findings of large anomalous Hall signal in nonferromagnetic and nonferrimagnetic materials suggest that the magnetization of the system is not a critical component for the realization of the anomalous Hall effect (AHE). Here, we present a combined theoretical and experimental study demonstrating the evolution of different mechanisms of AHE in a cubic Heusler system MnPt$_{1-x}$Ir$_x$Sn. With the help of magnetization and neutron diffraction studies, we show that the substitution of nonmagnetic Ir in place of Pt significantly reduces the net magnetic moment from 4.17 $ μ_B$/f.u. in MnPtSn to 2.78 $ μ_B$/f.u. for MnPt$_{0.5}$Ir$_{0.5}$Sn. In contrast, the anomalous Hall resistivity is enhanced by nearly three times from 1.6 $ μΩ$ cm in MnPtSn to about 5 $ μΩ$ cm for MnPt$_{0.5}$Ir$_{0.5}$Sn. The power law analysis of the Hall resistivity data suggests that the extrinsic contribution of AHE that dominates in the case of the parent MnPtSn almost vanishes for MnPt$_{0.5}$Ir$_{0.5}$Sn, where the intrinsic mechanism plays the major role. The experimental results are well supported by our theoretical study, which shows a considerable enhancement of the spin-orbit coupling when Ir is introduced into the system. Our finding of a crossover of the anomalous Hall effect with chemical engineering is a major contribution toward the recent interest in controlling the band topology of topological materials, both in bulk and thin-film forms.

cond-mat.mtrl-sci↗

arXiv2022

Serving Deep Learning Models with Deduplication from Relational Databases

There are significant benefits to serve deep learning models from relational databases. First, features extracted from databases do not need to be transferred to any decoupled deep learning systems for inferences, and thus the system management overhead can be significantly reduced. Second, in a relational database, data management along the storage hierarchy is fully integrated with query processing, and thus it can continue model serving even if the working set size exceeds the available memory. Applying model deduplication can greatly reduce the storage space, memory footprint, cache misses, and inference latency. However, existing data deduplication techniques are not applicable to the deep learning model serving applications in relational databases. They do not consider the impacts on model inference accuracy as well as the inconsistency between tensor blocks and database pages. This work proposed synergistic storage optimization techniques for duplication detection, page packing, and caching, to enhance database systems for model serving. We implemented the proposed approach in netsDB, an object-oriented relational database. Evaluation results show that our proposed techniques significantly improved the storage efficiency and the model inference latency, and serving models from relational databases outperformed existing deep learning frameworks when the working set size exceeds available memory.

cs.DB↗

arXiv2022

A Unified Cryptoprocessor for Lattice-based Signature and Key-exchange

We propose design methodologies for building a compact, unified and programmable cryptoprocessor architecture that computes post-quantum key agreement and digital signature. Synergies in the two types of cryptographic primitives are used to make the cryptoprocessor compact. As a case study, the cryptoprocessor architecture has been optimized targeting the signature scheme 'CRYSTALS-Dilithium' and the key encapsulation mechanism (KEM) 'Saber', both finalists in the NIST's post-quantum cryptography standardization project. The programmable cryptoprocessor executes key generations, encapsulations, decapsulations, signature generations, and signature verifications for all the security levels of Dilithium and Saber. On a Xilinx Ultrascale+ FPGA, the proposed cryptoprocessor consumes 18,406 LUTs, 9,323 FFs, 4 DSPs, and 24 BRAMs. It achieves 200 MHz clock frequency and finishes CCA-secure key-generation/encapsulation/decapsulation operations for LightSaber in 29.6/40.4/58.3$μ$s; for Saber in 54.9/69.7/94.9$μ$s; and for FireSaber in 87.6/108.0/139.4$μ$s, respectively. It finishes key-generation/sign/verify operations for Dilithium-2 in 70.9/151.6/75.2$μ$s; for Dilithium-3 in 114.7/237/127.6$μ$s; and for Dilithium-5 in 194.2/342.1/228.9$μ$s, respectively, for the best-case scenario. On UMC 65nm library for ASIC the latency is improved by a factor of two due to a 2x increase in clock frequency.

cs.CR↗

arXiv2022

Benchmark of DNN Model Search at Deployment Time

Deep learning has become the most popular direction in machine learning and artificial intelligence. However, the preparation of training data, as well as model training, are often time-consuming and become the bottleneck of the end-to-end machine learning lifecycle. Reusing models for inferring a dataset can avoid the costs of retraining. However, when there are multiple candidate models, it is challenging to discover the right model for reuse. Although there exist a number of model sharing platforms such as ModelDB, TensorFlow Hub, PyTorch Hub, and DLHub, most of these systems require model uploaders to manually specify the details of each model and model downloaders to screen keyword search results for selecting a model. We are lacking a highly productive model search tool that selects models for deployment without the need for any manual inspection and/or labeled data from the target domain. This paper proposes multiple model search strategies including various similarity-based approaches and non-similarity-based approaches. We design, implement, and evaluate these approaches on multiple model inference scenarios, including activity recognition, image recognition, text classification, natural language processing, and entity matching. The experimental evaluation showed that our proposed asymmetric similarity-based measurement, adaptivity, outperformed symmetric similarity-based measurements and non-similarity-based measurements in most of the workloads.

cs.IR↗

arXiv2021

Lachesis: Automatic Partitioning for UDF-Centric Analytics

Persistent partitioning is effective in avoiding expensive shuffling operations. However it remains a significant challenge to automate this process for Big Data analytics workloads that extensively use user defined functions (UDFs), where sub-computations are hard to be reused for partitionings compared to relational applications. In addition, functional dependency that is widely utilized for partitioning selection is often unavailable in the unstructured data that is ubiquitous in UDF-centric analytics. We propose the Lachesis system, which represents UDF-centric workloads as workflows of analyzable and reusable sub-computations. Lachesis further adopts a deep reinforcement learning model to infer which sub-computations should be used to partition the underlying data. This analysis is then applied to automatically optimize the storage of the data across applications to improve the performance and users' productivity.

cs.DB↗

arXiv2020

Survive the Schema Changes: Integration of Unmanaged Data Using Deep Learning

Data is the king in the age of AI. However data integration is often a laborious task that is hard to automate. Schema change is one significant obstacle to the automation of the end-to-end data integration process. Although there exist mechanisms such as query discovery and schema modification language to handle the problem, these approaches can only work with the assumption that the schema is maintained by a database. However, we observe diversified schema changes in heterogeneous data and open data, most of which has no schema defined. In this work, we propose to use deep learning to automatically deal with schema changes through a super cell representation and automatic injection of perturbations to the training data to make the model robust to schema changes. Our experimental results demonstrate that our proposed approach is effective for two real-world data integration scenarios: coronavirus data integration, and machine log integration.

cs.DB↗

arXiv2020

It's the Best Only When It Fits You Most: Finding Related Models for Serving Based on Dynamic Locality Sensitive Hashing

In recent, deep learning has become the most popular direction in machine learning and artificial intelligence. However, preparation of training data is often a bottleneck in the lifecycle of deploying a deep learning model for production or research. Reusing models for inferencing a dataset can greatly save the human costs required for training data creation. Although there exist a number of model sharing platform such as TensorFlow Hub, PyTorch Hub, DLHub, most of these systems require model uploaders to manually specify the details of each model and model downloaders to screen keyword search results for selecting a model. They are in lack of an automatic model searching tool. This paper proposes an end-to-end process of searching related models for serving based on the similarity of the target dataset and the training datasets of the available models. While there exist many similarity measurements, we study how to efficiently apply these metrics without pair-wise comparison and compare the effectiveness of these metrics. We find that our proposed adaptivity measurement which is based on Jensen-Shannon (JS) divergence, is an effective measurement, and its computation can be significantly accelerated by using the technique of locality sensitive hashing.

cs.IR↗

arXiv2019

Probing the multi spin-phonon coupling and local B-site disorder in Pr2CoFeO6 by Raman spectroscopy and correlation with its electronic structure by X-ray photoemission spectroscopy

Electronic structure near Fermi level of Pr2CoFeO6 (at 300 K) was investigated by X-ray photoemission spectroscopy (XPS) technique. All three cations, i.e., Pr, Co and Fe were found to be trivalent in nature. XPS analysis also suggested the system to be insulating in nature. Moreover, Raman spectroscopy study indicated the random distribution of the B-site ions (Co/Fe) triggered by same charge states. In temperature-dependent Raman study, the relative heights of the two observed phonon modes exhibited anomalous behaviour near magnetic transition temperature TN~270 K, thus indicating towards interplay between spin and phonon in the system. Furthermore, clear anomalous softening was observed below TN which confirmed the existence of strong spin-phonon coupling occurring for at least two phonon modes of the system. The line width analysis of the phonon modes essentially ruled out the role of magnetostriction effect in the observed phonon anomaly. The investigation of the lattice parameter variation across TN (obtained from the temperature-dependent neutron diffraction measurements) further confirmed the existence of the spin-phonon coupling.

cond-mat.str-el↗

arXiv2018

B-site spin-state and anti-site disorder driven multiple-magnetic phases: Griffiths phase, re-entrant cluster glass and exchange bias in double perovskite Pr$_2$CoFeO$_6$

We report the comprehensive experimental results identifying the magnetic spin ordering and the magnetization dynamics of a double perovskite Pr2CoFeO6 by employing the (dc and ac) magnetization, powder neutron diffraction (NPD) and X-ray magnetic circular dichroism (XMCD) techniques. X-ray diffraction and neutron diffraction studies revealed that Pr2CoFeO6 adopts a B-site disordered orthorhombic structure with space group Pnma. Additionally, ab initio band structure calculations performed on this system suggested an insulating anti-ferromagnetic (Fe-Fe) ground state. Magnetometry study showed the system to possess a spectrum of interesting magnetic phases including long range antiferromagnetic (canted) spin ordering (TN ~269 K), Griffiths phase, re-entrant cluster glass (RCG) (TG~ 34 K) and exchange bias. However, the NPD study divulged the exhibition of a long range G-type (below TN ~269 K) of spin ordering by Fe spins. Spin dynamics study by ac susceptibility technique confirmed the system possessing long range ordering at higher temperatureundergoes a RCG transition at ~34 K. Existence of Griffiths phase was confirmed by non-analytic field variation of magnetization and Heisenberg type temporal spin relaxation above long range ordering temperature TN ~269 K. The anti-site disorder related to the B-sites (Co/Fe) is found to be the main driving force forthe observed multiple magnetic phases. Furthermore, the electronic structure probed by the X-ray absorption spectroscopy (XAS) study suggested a nominal valance state of +3 for both of the B-site ions (Co/Fe) which in turn triggered the anti-site disorder in the system. Magnetic, XRD, NPD and XAS analysis yielded a low spin state (LS) for the Co3+ ions. The random non-magnetic dilution of magnetic Fe3+ (HS) ions by Co3+ (LS) ions essentially played a crucial role in manifesting the magnetic properties of the system.

cond-mat.mtrl-sci↗

arXiv2017

Field dependent neutron diffraction study in Ni50Mn38Sb12 Heusler alloy

In this paper, we present temperature and field dependent neutron diffraction (ND) study to unravel the structural and the magnetic properties in Ni50Mn38Sb12 Heusler system. This alloy shows martensitic transition from high temperature austenite cubic phase to low temperature martensite orthorhombic phase on cooling. At 3 K, the lattice parameters and magnetic moments are found to be almost insensitive to field. Just below the martensitic transition temperature, the martensite phase fraction is found to be 85%. Upon applying the field, the austenite phase becomes dominant, and the field induced reverse martensitic transition is clearly observed in the ND data. Therefore, the present study gives an estimate of the strength of the martensite phase or the sharpness of the martensitic transition. Variation of individual moments and the change in the phase fraction obtained from the analysis of the ND data vividly show the change in the magneto-structural state of the material across the transition.

cond-mat.mtrl-sci↗