The projects for cybersecurity include:
Differential Privacy for Graph Algorithms (Dana Dachman-Soled, University of Maryland)
Modeling Risk Triage in a Cyber-Compromised World (Emily Frye, The MITRE Corporation)
Targeted mobile digital forensics and privacy (Hongmei Chi, Florida A&M University)
The projects for climate change include:
Downscaling Climate Projections with Advanced Statistical Methods for Policy Decisions (Cecilia Bitz, University of Washington)
Decadal Prediction of the Climate and the Ocean: Advancing Understanding and Techniques (LuAnne Thompson, University of Washington)
Estimating the spatial extent of groundwater contamination in vulnerable communities in Los Angeles County (Michelle Miro, The RAND Corporation)
Differential Privacy for Graph Algorithms (Dana Dachman-Soled, University of Maryland)
Differential privacy formalizes what it means for a mechanism operating on data, collected from various individuals, to preserve the privacy of an individual. Our focus is on the use of third party cloud services for the storage of and computation on collected social network data, in the form of a graph. Datasets stored on the cloud can be automatically encrypted and tools from cryptography like homomorphic encryption and garbled circuits can be used to allow the cloud to compute on the encrypted data, without decrypting. However, even when the data itself is encrypted, access patterns (i.e., the sequence in which the nodes of the graph are accessed during the computation) can be used to allow the cloud to learn private information about individuals included in the dataset.
The goal of this project is to develop algorithms for data analytics on graphs. The algorithms will (1) be asymptotically efficient as the number of nodes grows, (2) preserve differential privacy through their access patterns, and (3) inform policy discussions about third party data warehousing. We anticipate using graph theory tools, such as node-edge-differential privacy, as well as probability theory tools, like optimal noise-adding mechanisms, in developing our algorithms.
*************************************************************************
Modeling Risk Triage in a Cyber-Compromised World (Emily Frye, The MITRE Corporation)
One of the greatest challenges facing the security of the United States is defining and mitigating the risk in an environment with ubiquitous connectivity. The endless pursuit of function and convenience, paired with the powerful marketing capability available to vendors today and enabled by policy choices that prioritize innovation, have resulted in a vast world of seamless yet opaque connections.
This project will bring together mathematicians and policymakers to address the question of where mathematical models and policy choices can be used to jointly understand and manage security situational awareness and inform decisions about least-cost placement of security burdens within a complex system of systems. Participants with a background or interests in stochastic analysis and statistical computing might be well suited for this project.
*************************************************************************
Targeted mobile digital forensics and privacy (Hongmei Chi, Florida A&M University)
With the rapid growth of mobile device usage, it is becoming increasingly urgent for law enforcement to seize and conduct forensic analysis on such devices when a crime occurs. The legal trend is to try to restrict the amount and type of data that can be analyzed such as through a warrant or appropriate consent, so an important issue is that law enforcement agencies acquire or extract only data relevant to the crime. Due to increasing concerns about privacy and the vast amount of information stored on smartphones, it is more and more challenging to conduct effective data extraction and analysis on such devices.
In this project, we will focus on discussing privacy and mobile digital forensics. We will learn various tools, such as OSForensics, SQLiteSpy and SimCOM, used in mobile digital forensics. We will explore how various encryption algorithms can protect user privacy on smartphones. In addition, we will show how we can obtain encrypted messages from third-party apps such as Whatsapp, Signal, etc. Participants will be conducting this analysis on openly available data as well as on their own data from their personal smartphones.
*************************************************************************
Downscaling Climate Projections with Advanced Statistical Methods for Policy Decisions (Cecilia Bitz, University of Washington)
Policy makers need information about future climate change on spatial scales much finer than is available from typical climate model grids. New and creative methods are being advanced to downscale climate change projections with statistical methods, and some simultaneously involve bias correction. Important requirements are to reliably downscale the climate quantity means, variability, extremes and trends, while also permitting uncertainty quantification. Neural networks meet these requirements by capturing linear and nonlinear relations, and at the same time they are analytically differentiable. Workshop participants will be invited to theorize about downscaling methods and to test their ideas on a case study for snowpack in the western United States. Datasets derived from both station data and fine-resolution climate models will be used to train and test statistical methods. Participants will discuss whether similar statistical methods could be used to develop new parameterizations for use in typical coarse-resolution climate models.
*************************************************************************
Decadal Prediction of the Climate and the Ocean: Advancing Understanding and Techniques (LuAnne Thompson, University of Washington)
Climate prediction one to 10 years in advance would be useful for many societal applications such as designing and investing in water resource infrastructure. Enormous progress has been made in seasonal climate prediction over the last several decades with the increased fidelity of these predictions relying on both improvements in the understanding and modeling of the tropical coupled ocean-atmosphere, and in the development of more sophisticated prediction methods.
The focus of most previous studies has been on sea surface temperature, ignoring the potential to extend prediction skill that could be obtained by considering the heat stored subsurface ocean. Over the last several decades, Argo profiling floats and satellite altimetry of sea levels have vastly increased the amount of information available on the subsurface ocean.
Here, we proposed to build a hierarchy of models and methods that can be used to examine the controls on the predictability of the state of the North Atlantic Ocean using both observations and climate model output. The simplest model for ocean-atmosphere interaction is a first order autoregressive model; however, this ignores ocean memory stored below the surface. We can increase complexity by adding additional predictands such as subsurface ocean temperature. However, the inherently coupled problem of climate prediction suggests that predictability is an emergent property of the system and that new approaches are needed that include knowledge of spatio-temporal statistical methods.
*************************************************************************
Estimating the spatial extent of groundwater contamination in vulnerable communities in Los Angeles County (Michelle Miro, The RAND Corporation)
Groundwater resources are increasingly drawn on as means to buffer surface water shortages during droughts as well as to improve the Los Angeles region’s reliance on local, rather than imported, water sources. However, the Los Angeles region is home to a legacy of contamination problems that threaten the quality and safety of groundwater as a drinking water resource. This project will examine the spatial extent of groundwater contamination in Los Angeles County’s groundwater basins and how this extent maps to vulnerable communities in the LA region. We will leverage multiple geospatial datasets and spatial interpolation and geostatistical methodologies, such as ordinary kriging and inverse distance weighting, to quantify the extent and magnitude of this contamination. We will then spatially join areas with estimated groundwater contamination to vulnerable communities that are dependent on groundwater resources for drinking water supply. The project will culminate in a hotspot map of drinking water systems and communities likely to be significantly impacted by groundwater contamination.
Copyright. All Rights Reserved.