The deployment of large-scale data analytics between on-premise and cloud sites, i.e., hybrid clouds, requires careful partitioning of both data and computation to avoid massive networking costs. We present Moirai, a cost-optimization framework that analyzes job accesses and data dependencies and optimizes the placement of both in hybrid clouds. Moirai informs the job scheduler of data location and access predictions, so it can determine where jobs should be executed to minimize data transfer costs. Our optimizer achieves scalability and cost efficiency by exploiting recurring jobs to identify data dependencies and job access characteristics and reduces the search space by excluding data not accessed recently.

An increasing demand for cross-cloud and cross-region data access is bringing forth challenges related to high data transfer costs and latency. In response, we introduce Macaron, an auto-configuring cache system designed to minimize cost for remote data access. A key insight behind Macaron is that cloud cache size is tied to cost, not hardware limits, shifting the way we think about cache design and eviction policies. Macaron dynamically configures cache size and utilizes a mix of cloud storage types to adapt to workload changes and reduce costs. We demonstrate that Macaron reduces cross-cloud workload costs by 65% and cross-region costs by 67%, mainly by reducing outgoing data transfer and by leveraging object storage alongside DRAM to reduce capacity cost.

In this paper, we address the black-box face reconstruction attack with two crucial requirements: query efficiency and generalizability. A practical attack must be query efficient due to limited access to the target black-box model, and the reconstructed face must be generalizable so it can be used to attack other face recognition systems. To this end, we propose a novel face reconstruction attack that optimizes the latent vector of a pre-trained StyleGAN generator. Unlike existing methods, our method is query efficient as neither training nor simultaneous updating of multiple latent vectors is required. Furthermore, we propose a simple initialization scheme that greatly enhances the generalizability of the proposed method. We demonstrate the effectiveness of our method by a thorough evaluation on LFW and CFP-FP datasets across multiple state-of-the-art face recognition models. Project Code: github.com/1ho0jin1/Black-box-Face-Reconstruction.

Smartphones have become indispensable in our lives, even for security-critical tasks. Traditional security measures such as PINs provide only one-time authentication, while biometrics enable continuous authentication in mobile devices. This paper introduces a simple, lightweight, pretrained Transformer dubbed PIEformer for open-set authentication (OSA) of multimodal touchstrokes and gait biometrics. Compared to conventional mobile closed-set authentication, OSA enables more secure and practical authentication, with genuine and impostor users disjoint from the training set. PIEFormer incorporates a novel implicit ensembling mechanism for extracting discriminative embeddings within an open-set environment and enhancing generalization performance. This approach learns multiple diverse sub-embeddings, capturing complementary aspects of biometrics data with minimal computational overhead, allowing Transformers to exhibit robust capabilities in OSA. Our proposed methods demonstrate state-of-the-art results on HMOG and BBMAS datasets, particularly in open-set scenarios compared to closed-set literature, thus bringing mobile biometric authentication closer to real-world applications.

Public cloud providers offer a diverse collection of storage types and configurations with different costs and performance SLAs. As a consequence, it is difficult to select the most cost-efficient allocations for storage backends, while satisfying a given workload's performance requirements, when moving data-heavy applications to the cloud. We present Mimir, a tool for automatically finding a cost-efficient virtual storage cluster configuration for a customer's storage workload and performance requirements. Importantly, Mimir considers all block storage types and configurations, and even heterogeneous mixes of them. In our experiments, compared to state-of-the-art approaches that consider only one storage type, Mimir finds configurations that reduce cost by up to 81% for real-application-based key-value store workloads.

In this paper, we focus on addressing the open-set face identification problem on a few-shot gallery by finetuning. The problem assumes a realistic scenario for face identification, where only a small number of face images is given for enrollment and any unknown identity must be rejected during identification. We observe that face recognition models pretrained on a large dataset and naively fine-tuned models perform poorly for this task. Motivated by this issue, we propose an effective fine-tuning scheme with classifier weight imprinting and exclusive BatchNorm layer tuning. For further improvement of rejection accuracy on unknown identities, we propose a novel matcher called Neighborhood Aware Cosine (NAC) that computes similarity based on neighborhood information. We validate the effectiveness of the proposed schemes thoroughly on large-scale face benchmarks across different convolutional neural network architectures. The source code for this project is available at: https://github.com/1ho0jin1/OSFI-by-FineTuning

The employment of high-performance servers and GPU accelerators for training deep neural network models have greatly accelerated recent advances in deep learning (DL). DL frameworks, such as TensorFlow, MXNet, and Caffe2, have emerged to assist DL researchers to train their models in a distributed manner. Although current DL frameworks scale well for image classification models, there remain opportunities for scalable distributed training on natural language processing (NLP) models. We found that current frameworks show relatively low scalability on training NLP models due to the lack of consideration to the difference in sparsity of model parameters. In this paper, we propose Parallax, a framework that optimizes data parallel training by utilizing the sparsity of model parameters. Parallax introduces a hybrid approach that combines Parameter Server and AllReduce architectures to optimize the amount of data transfer according to the sparsity. Experiments show that Parallax built atop Tensor-Flow achieves scalable training throughput on both dense and sparse models while requiring little effort from its users. Parallax achieves up to 2.8x, 6.02x speedup for NLP models than TensorFlow and Horovod with 48 GPUs, respectively. The training speed for the image classification models is equal to Horovod and 1.53x faster than TensorFlow.

The performance of distributed machine learning systems is dependent on their system configuration. However, configuring the system for optimal performance is challenging and time consuming even for experts due to the diverse runtime factors such as workloads or the system environment. We present cost-based optimization to automatically find a good system configuration for parameter server (PS) machine learning (ML) frameworks. We design and implement Cruise that applies the optimization technique to tune distributed PS ML execution automatically. Evaluation results on three ML applications verify that Cruise automates the system configuration of the applications to achieve good performance with minor reconfiguration costs.

This paper describes a simulation-based approach to establish whether a ring-oscillator always converges to the correct mode of operation regardless of its initial conditions and variability conditions. The verification is performed using a predictive global optimization algorithm that looks for a problematic initial state from a discretized state space. The algorithm explores the initial states that can maximize the settling time for the oscillator to reach its final steady state. If any of these initial states visited during the search is found exhibiting false oscillation behaviors for certain variability conditions, the initial state is reported as problematic. On the other hand, if the initial state with the globally maximum settling time is found without discovering such problematic states, the oscillator is reported free of start-up failures. It can be shown that despite the finite number of initial state candidates considered and finite number of Monte-Carlo samples to model variability, the proposed algorithm can verify the oscillator to a prescribed confidence level. Demonstrated on various even-stage differential ring oscillators, the algorithm was able to validate the circuit for 99% yield with 99.9% confidence level by evaluating 7~60 initial states each with 1,000 Monte-Carlo samples. To our knowledge, this is the first algorithm ever reported to address start-up failures with variability.

A cascaded hybrid phase-locked loop (PLL) fabricated in a 65-nm CMOS process consumes 21 mW and occupies 0.4 mm2. An all-digital PLL (ADPLL) with piecewise linear calibrated hierarchical time-to-digital converter is proposed to achieve a wide operation range, and a charge-pump PLL (CPPLL) with an auxiliary (AUX) charge-pump for low current mismatch is cascaded to filter out the ADPLL output noise. The ADPLL achieves low long-term jitter regardless of the leakage current, and the CPPLL realizes low short-term jitter using a self-biased technique and the AUX charge pump. A phase-selectable divider is also proposed to divide the clock frequency while keeping the relative phase difference constant. The measured peak-to-peak short-term and long-term jitters at an output frequency of 315 MHz are 40 and 70 pspp, respectively, with a multiplication factor of 1024.