Password-based user authentication has been pervasively used and rigorously studied for many years, while face biometrics based user authentication has become increasingly popular recently. In this talk, I will answer why it is so difficult to make user authentication highly secure and highly usable at the same time based on password only. In addition, I will answer how we can push the usability of password-based two-factor authentication to its limit, and how we can make face authentication highly secure without sacrificing much of its high usability.
Ever since the beginning of mobile data in the early days of 2G and 2.5G networks, we've fundamentally built mobile and even IoT applications the same way – fetching data from the Cloud or other central source via a REST-like protocol. With each new generation of mobile networks, we argued that the network is getting faster, the bandwidth is increasing, and therefore network updates will address any and all user experience issues. However, mobile devices and IoT devices that change position in space fundamentally experience regular and often rapid changes in bandwidth, connectivity, and l
The first milestone in the CIS Ph.D. program is the Directed Research Project (DRP). Each student must complete this milestone within their first two years in the program. Typically, students devote the summer after their first year in the program exclusively to their research projects. In this colloquium, seven of our Department's Ph.D. students will present summaries on the progress on their DRP's over this past summer.
I will present some work around the OpenSHMEM interface. At first I will present POSH, an implementation of the OpenSHMEM standard that aims at using a software layer as thin as possible, using techniques including metaprogramming. Then I will present results issued from a collaboration with colleagues who work on model checking, a formal verification technique that consists in exploring all the possible states of a system in order to verify properties on this system. The combinatorial explosion of the number of states makes model checking both computation-intensive and memory-intensive.
A growing disparity between supercomputer computation speeds and I/O rates makes it increasingly infeasible for applications to save all results for offline analysis. Instead, applications must analyze and reduce data in situ so as to output only those results needed to answer target scientific question(s).
All applications on the planet are in or moving in to the Cloud, supporting what is important and critical to our lives and our work. I will talk about how we can build the physical and virtual network infrastructure to enable high velocity in innovation along with high scale and reliability, with fundamental components open or open source, available for you to pick up from github.
I am looking forward to meeting and talking with faculty and students.
The deceleration of transistor feature size scaling has motivated growing adoption of specialized accelerators implemented as GPUs, FPGAs, ASICs, and more recently new types of computing such as neuromorphic, bio-inspired, ultra-low energy, reversible, stochastic, optical, quantum, combinations, and others unforeseen. There is a tension between specialization and generalization, with the current state trending to master-slave models where accelerators (slaves) are instructed by a general purpose system (master) running an Operating System (OS).
Classes of intersection and contact graphs of geometric objects are well studied both for their relevance to graph visualization and for their interesting structural and algorithmic properties. Many of them can be recognized in polynomial time (e.g., interval graphs, circle graphs, permutation graphs and many more). The aim of the talk is to compare two natural generalizations of the recognition problem: RepExt (when a part of the input graph comes pre-represented) and SimRep (when two or more input graphs are to be represented simultaneously, i.e.
Recent advances in computer vision and the proliferation of video cameras in wearables and IoT devices have created the potential for a large number of novel applications. These same devices and algorithms, though, raise significant privacy concerns, as we can now capture and process enormous amounts of video data. One solution for this is to store only privacy-preserving video, which has been pre-processed to remove identifying information.
To design adaptive and personalized learning system, it is crucial to have better understandings of students' behavior and performance. This talk will introduce three studies on quantifying .students' behavior and performance in different online educational settings by applying machine learning and natural language processing techniques.