No abstract

Personal data garnered from various sensors are often offloaded by applications to the cloud for analytics. This leads to a potential risk of disclosing private user information. We observe that the analytics run on the cloud are often limited to a machine learning model such as predicting a user’s activity using an activity classifier. We present Olympus, a privacy framework that limits the risk of disclosing private user information by obfuscating sensor data while minimally affecting the functionality the data are intended for. Olympus achieves privacy by designing a utility aware obfuscation mechanism, where privacy and utility requirements are modeled as adversarial networks. By rigorous and comprehensive evaluation on a real world app and on benchmark datasets, we show that Olympus successfully limits the disclosure of private information without significantly affecting functionality of the application.

The Genomic Standards Consortium (GSC) invited a representative of the Long-Term Ecological Research (LTER) to its fifth workshop to present the Ecological Metadata Language (EML) metadata standard and its relationship to the Minimum Information about a Genome/Metagenome Sequence (MIGS/MIMS) and its implementation, the Genomic Contextual Data Markup Language (GCDML). The LTER is one of the top National Science Foundation (NSF) programs in biology since 1980, representing diverse ecosystems and creating long-term, interdisciplinary research, synthesis of information, and theory. The adoption of EML as the LTER network standard has been key to build network synthesis architectures based on high-quality standardized metadata. EML is the NSF-recognized metadata standard for LTER, and EML is a criteria used to review the LTER program progress. At the workshop, a potential crosswalk between the GCDML and EML was explored. Also, collaboration between the LTER and GSC developers was proposed to join efforts toward a common metadata cataloging designer's tool. The community adoption success of a metadata standard depends, among other factors, on the tools and trainings developed to use the standard. LTER's experience in embracing EML may help GSC to achieve similar success. A possible collaboration between LTER and GSC to provide training opportunities for GCDML and the associated tools is being explored. Finally, LTER is investigating EML enhancements to better accommodate genomics data, possibly integrating the GCDML schema into EML. All these action items have been accepted by the LTER contingent, and further collaboration between the GSC and LTER is expected.

Datacenter-scale AI training clusters consisting of thousands of domain-specific accelerators (DSA) are used to train increasinglycomplex deep learning models. These clusters rely on a data storage and ingestion (DSI) pipeline, responsible for storing exabytes of training data and serving it at tens of terabytes per second. As DSAs continue to push training efficiency and throughput, the DSI pipeline is becoming the dominating factor that constrains the overall training performance and capacity. Innovations that improve the efficiency and performance of DSI systems and hardware are urgent, demanding a deep understanding of DSI characteristics and infrastructure at scale. This paper presents Meta's end-to-end DSI pipeline, composed of a central data warehouse built on distributed storage and a Data PreProcessing Service that scales to eliminate data stalls. We characterize how hundreds of models are collaboratively trained across geo-distributed datacenters via diverse and continuous training jobs. These training jobs read and heavily filter massive and evolving datasets, resulting in popular features and samples used across training jobs. We measure the intense network, memory, and compute resources required by each training job to preprocess samples during training. Finally, we synthesize key takeaways based on our production infrastructure characterization. These include identifying hardware bottlenecks, discussing opportunities for heterogeneous DSI hardware, motivating research in datacenter scheduling and benchmark datasets, and assimilating lessons learned in optimizing DSI infrastructure. CCS CONCEPTS• Software and its engineering → Distributed systems organizing principles; • Information systems → Database management system engines; • Computing methodologies → Machine learning.

Remote sensing data from satellites have provided valuable information on the state of the earth for several decades.