arXiv subjects

Junsoo Kim

Publications and source records attributed to Junsoo Kim.

33 records · Page 2Linked to original sources

arXiv2024

LPU: A Latency-Optimized and Highly Scalable Processor for Large Language Model Inference

The explosive arrival of OpenAI's ChatGPT has fueled the globalization of large language model (LLM), which consists of billions of pretrained parameters that embodies the aspects of syntax and semantics. HyperAccel introduces latency processing unit (LPU), a latency-optimized and highly scalable processor architecture for the acceleration of LLM inference. LPU perfectly balances the memory bandwidth and compute logic with streamlined dataflow to maximize performance and efficiency. LPU is equipped with expandable synchronization link (ESL) that hides data synchronization latency between multiple LPUs. HyperDex complements LPU as an intuitive software framework to run LLM applications. LPU achieves 1.25 ms/token and 20.9 ms/token for 1.3B and 66B model, respectively, which is 2.09x and 1.37x faster than the GPU. LPU, synthesized using Samsung 4nm process, has total area of 0.824 mm2 and power consumption of 284.31 mW. LPU-based servers achieve 1.33x and 1.32x energy efficiency over NVIDIA H100 and L4 servers, respectively.

cs.AR↗

arXiv2024

A code-driven tutorial on encrypted control: From pioneering realizations to modern implementations

The growing interconnectivity in control systems due to robust wireless communication and cloud usage paves the way for exciting new opportunities such as data-driven control and service-based decision-making. At the same time, connected systems are susceptible to cyberattacks and data leakages. Against this background, encrypted control aims to increase the security and safety of cyber-physical systems. A central goal is to ensure confidentiality of process data during networked controller evaluations, which is enabled by, e.g., homomorphic encryption. However, the integration of advanced cryptographic systems renders the design of encrypted controllers an interdisciplinary challenge. This code-driven tutorial paper aims to facilitate the access to encrypted control by providing exemplary realizations based on popular homomorphic cryptosystems. In particular, we discuss the encrypted implementation of state feedback and PI controllers using the Paillier, GSW, and CKKS cryptosystem.

eess.SY↗

arXiv2023

Further methods for encrypted linear dynamic controllers utilizing re-encryption

Homomorphic encryption (HE) applied to a networked controller enables secure operation, but in most cases it allows for addition and multiplication over integers only, because of computation efficiency. Several related results deal with such constraints by means of re-encrypted controller output, based on which the controller can be re-constructed and operate over integers. This paper presents two modified methods based on the output re-encryption, which will further reduce the required communication effort and computational burden, respectively.

eess.SY↗

arXiv2023

Complexity reduction for resilient state estimation of uniformly observable nonlinear systems

A resilient state estimation scheme for uniformly observable nonlinear systems, based on a method for local identification of sensor attacks, is presented. The estimation problem is combinatorial in nature, and so many methods require substantial computational and storage resources as the number of sensors increases. To reduce the complexity, the proposed method performs the attack identification with local subsets of the measurements, not with the set of all measurements. A condition for nonlinear attack identification is introduced as a relaxed version of existing redundant observability condition. It is shown that an attack identification can be performed even when the state cannot be recovered from the measurements. As a result, although a portion of measurements are compromised, they can be locally identified and excluded from the state estimation, and thus the true state can be recovered. Simulation results demonstrate the effectiveness of the proposed scheme.

eess.SY↗

arXiv2022

Comparison of encrypted control approaches and tutorial on dynamic systems using LWE-based homomorphic encryption

Encrypted control has been introduced to protect controller data by encryption at the stage of computation and communication, by performing the computation directly on encrypted data. In this article, we first review and categorize recent relevant studies on encrypted control. Approaches based on homomorphic encryption, multi-party computation, and secret sharing are introduced, compared, and then discussed with respect to computational complexity, communication load, enabled operations, security, and research directions. We proceed to discuss a current challenge in the application of homomorphic encryption to dynamic systems, where arithmetic operations other than integer addition and multiplication are limited. We also introduce a homomorphic cryptosystem called ``GSW-LWE'' and discuss its benefits that allow for recursive multiplication of encrypted dynamic systems, without use of computationally expensive bootstrapping techniques.

cs.CR↗

arXiv2022

DFX: A Low-latency Multi-FPGA Appliance for Accelerating Transformer-based Text Generation

Transformer is a deep learning language model widely used for natural language processing (NLP) services in datacenters. Among transformer models, Generative Pre-trained Transformer (GPT) has achieved remarkable performance in text generation, or natural language generation (NLG), which needs the processing of a large input context in the summarization stage, followed by the generation stage that produces a single word at a time. The conventional platforms such as GPU are specialized for the parallel processing of large inputs in the summarization stage, but their performance significantly degrades in the generation stage due to its sequential characteristic. Therefore, an efficient hardware platform is required to address the high latency caused by the sequential characteristic of text generation. In this paper, we present DFX, a multi-FPGA acceleration appliance that executes GPT-2 model inference end-to-end with low latency and high throughput in both summarization and generation stages. DFX uses model parallelism and optimized dataflow that is model-and-hardware-aware for fast simultaneous workload execution among devices. Its compute cores operate on custom instructions and provide GPT-2 operations end-to-end. We implement the proposed hardware architecture on four Xilinx Alveo U280 FPGAs and utilize all of the channels of the high bandwidth memory (HBM) and the maximum number of compute resources for high hardware efficiency. DFX achieves 5.58x speedup and 3.99x energy efficiency over four NVIDIA V100 GPUs on the modern GPT-2 model. DFX is also 8.21x more cost-effective than the GPU appliance, suggesting that it is a promising solution for text generation workloads in cloud datacenters.

eess.SY↗

arXiv2021

Dynamic controller that operates over homomorphically encrypted data for infinite time horizon

In this paper, we present a dynamic feedback controller that computes the next state and the control signal over encrypted data using homomorphic properties of cryptosystems, whose performance is equivalent to the linear dynamic controllers over real-valued data. Assuming that the input as well as the output of the plant is encrypted and transmitted back to the controller, it is shown that the state matrix of any linear time-invariant controller can be always converted to a matrix of integer components. This allows the dynamic feedback controller to operate for infinite time horizon without decryption or reset of its internal state. For implementation in practice, we illustrate the use of a cryptosystem that is based on the Learning With Errors problem, which allows both multiplication and addition over encrypted data. It is also shown that the effect of injected random numbers during encryption for security can be maintained within a small bound by way of the closed-loop stability.

eess.SY↗

arXiv2021

Toward nonlinear dynamic control over encrypted data for infinite time horizon

Recent studies on encrypted control using homomorphic encryption allow secure operation by directly performing computations on encrypted data without decryption. Implementing dynamic controllers on encrypted data presents unique challenges due to limitations on the number of operations on an encrypted message. Hence, it may not be possible to perform the recursive operations for an infinite time horizon. In this note, we demonstrate that it is possible to run a dynamic controller over encrypted data for an infinite time horizon if the output of the controller can be represented as a function of a fixed number of previous inputs and outputs. The presented implementation requires encryption at both input and output of the plant. We identify a class of nonlinear systems that can accommodate the proposed implementation. The closed-loop performance can be guaranteed using the proposed encrypted controller by ensuring that quantization error is made arbitrarily small with appropriate choice of parameters. We show that the proposed method is amenable to linear systems (as a subset of the said nonlinear systems) with performance guarantees.

eess.SY↗

arXiv2021

Homomorphic Encryption-Enabled Distance-Based Distributed Formation Control with Distance Mismatch Estimators

This paper considers the use of fully homomorphic encryption for the realisation of distributed formation control of multi-agent systems via edge computer. In our proposed framework, the distributed control computation in the edge computer uses only the encrypted data without the need for a reset mechanism that is commonly required to avoid error accumulation. Simulation results show that, despite the use of encrypted data on the controller and errors introduced by the quantization process prior to the encryption, the formation is able to converge to the desired shape. The proposed architecture offers insight on the mechanism for realising distributed control computation in an edge/cloud computer while preserving the privacy of local information coming from each agent.

eess.SY↗

arXiv2021

Fully Distributed Resilient State Estimation based on Distributed Median Solver

In this paper, we present a scheme of fully distributed resilient state estimation for linear dynamical systems under sensor attacks. The proposed state observer consists of a network of local observers, where each of them utilizes local measurements and information transmitted from the neighbors. As a fully distributed scheme, it does not necessarily collect a majority of sensing data for the sake of attack identification, while the compromised sensors are eventually identified by the distributed network and excluded from the observers. For this, the overall network (not the individual local observer) is assumed to have redundant sensors and assumed to be connected. The proposed scheme is based on a novel design of a distributed median solver, which approximately recovers the median value of local estimates.

eess.SY↗

arXiv2021

LaPred: Lane-Aware Prediction of Multi-Modal Future Trajectories of Dynamic Agents

In this paper, we address the problem of predicting the future motion of a dynamic agent (called a target agent) given its current and past states as well as the information on its environment. It is paramount to develop a prediction model that can exploit the contextual information in both static and dynamic environments surrounding the target agent and generate diverse trajectory samples that are meaningful in a traffic context. We propose a novel prediction model, referred to as the lane-aware prediction (LaPred) network, which uses the instance-level lane entities extracted from a semantic map to predict the multi-modal future trajectories. For each lane candidate found in the neighborhood of the target agent, LaPred extracts the joint features relating the lane and the trajectories of the neighboring agents. Then, the features for all lane candidates are fused with the attention weights learned through a self-supervised learning task that identifies the lane candidate likely to be followed by the target agent. Using the instance-level lane information, LaPred can produce the trajectories compliant with the surroundings better than 2D raster image-based methods and generate the diverse future trajectories given multiple lane candidates. The experiments conducted on the public nuScenes dataset and Argoverse dataset demonstrate that the proposed LaPred method significantly outperforms the existing prediction models, achieving state-of-the-art performance in the benchmarks.

cs.CV↗

arXiv2021

Electric-magnetic duality as a quantum operator and more symmetries of $U(1)$ gauge theory

We promote the Noether charge of the electric-magnetic duality symmetry of $U(1)$ gauge theory, "$G$" to a quantum operator. We construct ladder operators, $D_{(\pm)a}^\dagger(k)$ and $D_{(\pm)a}(k)$ which create and annihilate the simultaneous quantum eigen states of the quantum Hamiltonian(or number) and the electric-magnetic duality operators respectively. Therefore all the quantum states of the $U(1)$ gauge fields can be expressed by a form of $|E,g\rangle$, where $E$ is the energy of the state, the $g$ is the eigen value of the quantum operator $G$, where the $g$ is quantized in the unit of 1. We also show that 10 independent bilinears comprised of the creation and annihilation operators can form $SO(2,3)$ which is as demonstrated in the Dirac's paper published in 1962. The number operator and the electric-magnetic duality operator are the members of the $SO(2,3)$ generators. We note that there are two more generators which commute with the number operator(or Hamiltonian). We prove that these generators are indeed symmetries of the $U(1)$ gauge field theory action.

hep-th↗

arXiv2019

Comprehensive Introduction to Fully Homomorphic Encryption for Dynamic Feedback Controller via LWE-based Cryptosystem

The cryptosystem based on the Learning-with-Errors (LWE) problem is considered as a post-quantum cryptosystem, because it is not based on the factoring problem with large primes which is easily solved by a quantum computer. Moreover, the LWE-based cryptosystem allows fully homomorphic arithmetics so that two encrypted variables can be added and multiplied without decrypting them. This chapter provides a comprehensive introduction to the LWE-based cryptosystem with examples. A key to the security of the LWE-based cryptosystem is the injection of random errors in the ciphertexts, which however hinders unlimited recursive operation of homomorphic arithmetics on ciphertexts due to the growth of the error. We show that this limitation can be overcome when the cryptosystem is used for a dynamic feedback controller that guarantees stability of the closed-loop system. Finally, we illustrate through MATLAB codes how the LWE-based cryptosystem can be customized to build a secure feedback control system. This chapter is written for the control engineers who do not have background on cryptosystems.

eess.SY↗

arXiv2018

Detection of Sensor Attack and Resilient State Estimation for Uniformly Observable Nonlinear Systems having Redundant Sensors

This paper presents a detection algorithm for sensor attacks and a resilient state estimation scheme for a class of uniformly observable nonlinear systems. An adversary is supposed to corrupt a subset of sensors with the possibly unbounded signals, while the system has sensor redundancy. We design an individual high-gain observer for each measurement output so that only the observable portion of the system state is obtained. Then, a nonlinear error correcting problem is solved by collecting all the information from those partial observers and exploiting redundancy. A computationally efficient, on-line monitoring scheme is presented for attack detection. Based on the attack detection scheme, an algorithm for resilient state estimation is provided. The simulation results demonstrate the effectiveness of the proposed algorithm.

eess.SY↗

arXiv2016

Optical Characterization of PtSi/Si by Spectroscopic Ellipsometry

We report optical characterization of PtSi films for thermoelectric device applications by nondestructive spectroscopic ellipsometry (SE). Pt monolayer and Pt-Si multilayer which consists of 3 pairs of Pt and Si layers were deposited on p-doped-silicon substrates by sputtering method and then rapid annealing process was done to form PtSi films through intermixing of Pt and Si atoms at the interface. Pseudodielectric function data <ε> = <ε1> + i<ε2> of the PtSi/Si samples were obtained from 1.12 to 6.52 eV by using spectroscopic ellipsometry. Employing Tauc-Lorentz and Drude models, the dielectric function (ε) of PtSi films were determined. We found that the composition ratio of Pt:Si is nearly 1:1 for PtSi monolayer and we observed transitions between occupied and unoccupied states in Pt 5d states. We also observed formation of PtSi layers in Pt-Si multilayer sample. The SE results were confirmed by the transmission electron microscopy and energy dispersive X-ray spectroscopy.

cond-mat.mtrl-sci↗