No abstract

Forecasting natural disasters such as inundations can be of great help for emergency bodies and first responders. In coastal communities, this risk is often associated with storm surge. To produce flood forecasts for coastal communities, a system must incorporate models capable of simulating such events based on forecasted weather conditions. In this work, a system for forecasting inundations based predominantly on storm surge is explored. An automation and a coupling strategy were implemented to produce forecasted flood maps automatically. The system leverages an ocean circulation model and a channel water flow model to estimate flood events in South Texas specially alongside the Lower Laguna Madre. The system around the models is implemented using Python and the meteorological forcing input is obtained from weather forecasting models maintained by the National Oceanic and Atmospheric Administration. The forecasted weather data retrieval, data processing and automation of the models are successful, and the complete stack of software can be deployed locally or in cloud solutions to accelerate computations. The resulting system performs as expected and successfully produces flood maps automatically providing vital information for flood emergency management in coastal communities.

Recommender systems are widely used in a variety of scenarios, including online shopping, social network, and contents distribution. As users rely more on recommender systems for information retrieval, they also become attractive targets for cyber-attacks. The high-level idea of attacking a recommender system is straightforward. An adversary selects a strategy to inject manipulated data into the database of the recommender system to influence the recommendation results, which is also known as a profile injection attack. Most existing works treat attacking and protection in a static manner, i.e., they only consider the adversary's behavior when analyzing the influence without considering normal users' activities. However, most recommender systems have a large number of normal users who also add data to the database, the effects of which are largely ignored when considering the protection of a recommender system. We take normal users' contributions into consideration and analyze popular attacks against a recommender system. We also propose a general protection framework under this dynamic setting.