January 23 Organizational meeting
Organizational meeting: Sign up to give a talk!
January 30 Ali Mashtizadeh
Abstract: Modern users want instant access to their data on any machine they log into, offline access when they are traveling, and file history to access accidentally deleted or damaged data. We built Ori, a file system that supports all these use cases, showing how a single file system can be suitable for multiple different configurations: as a general purpose local file system, as a distributed one, and as one that supports versioning. The key insight is that version control and network file systems go hand in hand: replication and versioning are needed for offline use and later merging/conflict resolution, and network file system-like mechanisms to instantly access remote files are needed to make general use practical. The latter is achieved through InstaClone, a novel mechanism introduced by Ori. Ori runs on Mac OS, Linux, and FreeBSD, and show a small overhead when used as a general purpose file system, while bringing the benefits of versioning and instant remote access to all files.
February 6 Vimal Jeyakumar
Title: EyeQ: Protecting your Network Performance
Abstract: Today, a datacentre infrastructure provider (e.g. Amazon AWS, Windows Azure) hosts diverse applications and not all of them can be trusted. While "virtualisation" has made significant advances in isolating CPU performance, there's little to no protection for network bandwidth. Contention occurring at timescales of a few milliseconds, invisible on human timescales, can degrade long term performance. This talk is about one practical approach to "virtualise" network bandwidth even in the presence of adversarial traffic patterns.
February 13 Edward Yang
Abstract: To ensure the confidentiality and integrity of web content, modern web browsers enforce isolation between content and scripts from different domains with the same-origin policy. However, many web applications require cross-origin sharing of code and data. This conflict between isolation and sharing has led to an ad hoc implementation of the SOP that has proven vulnerable to such attacks as cross-site scripting, cross-site request forgery, and browser privacy leaks. In this talk, we argue that information flow control (IFC) subsumes same-origin policy. We'll show how to express existing browser policies in IFC and show that IFC can also help developers build complex sites such as mashups, which are notoriously difficult to implement securely under the SOP.
February 20 Amit Levy
February 27 Eric Lam
Title: A Learning Theoretic Approach to Non-Interactive Database Privacy by Blum, Ligett, and Roth, and application of differential privacy to healthcare data
Summary: I will present some of the results and techniques employed in the STOC 2008 paper by Blum, et al. Given a database containing sensitive information, the authors applied learning theory to create a synthetic database that can answer queries from a given concept class while providing non-trivial utility for each of the queries, and preserving privacy. This paper circumvented some earlier negative results which state that a privacy-preserving database access mechanism cannot answer more than a sublinear number of queries accurately, by only guaranteeing usefulness for queries in restricted classes.
I will also present some of the critiques by healthcare professionals on the application of differential privacy mechanisms to release sensitive healthcare data.
March 6 Ananth Raghunathan
Title: Randomness Extractors in Cryptography
Summary: In this talk, I will cover several applications of randomness extractors in cryptography. This talk is motivated by the Leftover Hash Lemma (which briefly states that a family of pairwise independent hash functions is a good randomness extractor) and its several variants. They have applications in cryptography spanning amplification of security, entropic security, deterministic extraction, and deterministic encryption schemes. I will discuss the Leftover Hash Lemma, a couple of its variants, and how they apply to the topics mentioned above and conclude with the state of the art on the Leftover Hash Lemma and open problems.