An official website of the United States government
Homomorphic encryption (HE) algorithms, particularly the Cheon-Kim-Kim-Song (CKKS) scheme, offer significant potential for secure computation on encrypted data, making them valuable for privacy-preserving machine learning. However, high latency in large integer operations in the CKKS algorithm hinders the processing of large datasets and complex computations. This paper proposes a novel strategy that combines lossless data compression techniques with the parallel processing power of graphics processing units to address these challenges. Our approach demonstrably reduces data size by 90% and achieves significant speedups of up to 100 times compared to conventional approaches. This method ensures data confidentiality while mitigating performance bottlenecks in CKKS-based computations, paving the way for more efficient and scalable HE applications.
more »
« less
Massively Scalable Parallel KMeans on the HPCC Systems Platform
Clustering algorithms are an important part of unsupervised machine learning. With Big Data, applying clustering algorithms such as KMeans has become a challenge due to the significantly larger volume of data and the computational complexity of the standard approach, Lloyd's algorithm. This work aims to tackle this challenge by transforming the classic clustering KMeans algorithm to be highly scalable and to be able to operate on Big Data. We leverage the distributed computing environment of the HPCC Systems platform. The presented KMeans algorithm adopts a hybrid parallelism method to achieve a massively scalable parallel KMeans. Our approach can save a significant amount of time of researchers and machine learning practitioners who train hundreds of models on a daily basis. The performance is evaluated with different size datasets and clusters and the results show a significant scalabilty of the scalable parallel KMeans algorithm.
more »
« less
- Award ID(s):
- 1725573
- PAR ID:
- 10201358
- Date Published:
- Journal Name:
- 2019 4th International Conference on Computational Systems and Information Technology for Sustainable Solution
- Page Range / eLocation ID:
- 1 to 8
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Hierarchical clustering is a fundamental tool in data mining, machine learning and statistics. Popular hierarchical clustering algorithms include top-down divisive approaches such as bisecting k-means, k-median, and k-center and bottom-up agglomerative approaches such as single- linkage, average-linkage, and centroid-linkage. Unfortunately, only a few scalable hierarchical clustering algorithms are known, mostly based on the single-linkage algorithm. So, as datasets increase in size every day, there is a pressing need to scale other popular methods. We introduce efficient distributed algorithms for bisecting k-means, k- median, and k-center as well as centroid-linkage. In particular, we first formalize a notion of closeness for a hierarchical clustering algorithm, and then we use this notion to design new scalable distributed methods with strong worst case bounds on the running time and the quality of the solutions. Finally, we show experimentally that the introduced algorithms are efficient and close to their sequential variants in practice.more » « less
-
An increasing number of systems are being designed by gathering significant amounts of data and then optimizing the system parameters directly using the obtained data. Often this is done without analyzing the dataset structure. As task complexity, data size, and parameters all increase to millions or even billions, data summarization is becoming a major challenge. In this work, we investigate data summarization via dictionary learning (DL), leveraging the properties of recently introduced non-negative kernel regression (NNK) graphs. Our proposed NNK-Means, unlike previous DL techniques, such as kSVD, learns geometric dictionaries with atoms that are representative of the input data space. Experiments show that summarization using NNK-Means can provide better class separation compared to linear and kernel versions of kMeans and kSVD. Moreover, NNK-Means is scalable, with runtime complexity similar to that of kMeans.more » « less
-
In this paper, we propose a responsive autonomic and data-driven adaptive virtual networking framework (RAvN) to detect and mitigate anomalous network behavior. The proposed detection scheme detects both low rate and high rate denial of service (DoS) attacks using (1) a new Centroid-based clustering technique, (2) a proposed Intragroup variance technique for data features within network traffic (C.Intra) and (3) a multivariate Gaussian distribution model fitted to the constant changes in the IP addresses of the network. RAvN integrates the adaptive reconfigurable features of a popular SDN platform (open networking operating system (ONOS)); the network performance statistics provided by traffic monitoring tools (such as T-shark or sflow-RT); and the analytics and decision-making tools provided by new and current machine learning techniques. The decision making and execution components generate adaptive policy updates (i.e. anomalous mitigation solutions) on-the-fly to the ONOS SDN controller for updating network configurations and flows. In addition, we compare our anomaly detection schemes for detecting low rate and high rate DoS attacks versus a commonly used unsupervised machine learning technique, Kmeans. Kmeans recorded 72.38% accuracy, while the multivariate clustering and the Intra-group variance methods recorded 80.54% and 96.13% accuracy respectively, a significant performance improvement.more » « less
-
The push to automate and digitize the electric grid has led to widespread installation of Phasor Measurement Units (PMUs) for improved real-time wide-area system monitoring and control. Nevertheless, transforming large volumes of highresolution PMU measurements into actionable insights remains challenging. A central challenge is creating flexible and scalable online anomaly detection in PMU data streams. PMU data can hold multiple types of anomalies arising in the physical system or the cyber system (measurements and communication networks). Increasing the grid situational awareness for noisy measurement data and Bad Data (BD) anomalies has become more and more significant. Number of machine learning, data analytics and physics based algorithms have been developed for anomaly detection, but need to be validated with realistic synchophasor data. Access to field data is very challenging due to confidentiality and security reasons. This paper presents a method for generating realistic synchrophasor data for the given synthetic network as well as event and bad data detection and classification algorithms. The developed algorithms include Bayesian and change-point techniques to identify anomalies, a statistical approach for event localization and multi-step clustering approach for event classification. Developed algorithms have been validated with satisfactory results for multiple examples of power system events including faults and load/generator/capacitor variations/switching for an IEEE test system. Set of synchrophasor data will be available publicly for other researchers.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/CSITSS47250.2019.9031047
