October 3, 2007
The ubiquity of sensitive information in today’s computer systems combined with the increasing connectivity of these systems require security controls to ensure information is not leaked to unauthorized locations. To this end, information flow security mechanisms enforce application-level security by tracking and controlling the dissemination of information through programs. Though much foundational work has been done in this area, current systems are hampered by a high learning curve, many code annotations, poorly defined security policies, and imprecision in the analysis.
In this talk, I will describe a static information flow type inference system for Java. The focus of the system is on specifying security policies of IO channels only, making code annotations unnecessary, as the type system automatically infers all of the security labels throughout the code. The type system includes a high degree of polymorphism, necessary to allow classes to be used in multiple security contexts, and to properly distinguish the security policies of different IO channels. Top-level policy specifications allow the policy to be clearly defined and visible in the API. I will conclude by describing how the type system is proved correct with a noninterference result: that high security data is not visible to a low security observer.
October 26, 2007
In complexity theory based cryptography, the goal is to devise cryptographic schemes which can be proven secure (under some well defined notion of security) using some assumption about the intractability of a computational problem. Usually this is done via a reduction that given an hypothetical attack to the cryptographic scheme builds an algorithm to solve the assumed intractable problem. The ultimate goal is to find the most efficient scheme, which satisfies a very strong definition of security, under the weakest possible computational assumption. In the past 25 years, complexity theory based cryptography has produced several impressive results, at least under the two last measures: very strong notions of security for tasks like encryption, signatures, psedorandom generation etc. can be achieved under minimal computational assumptions, even though often (but not always) by schemes that are too inefficient to be used in practice. Bridging this gap between “provable security” and “efficiency” has been a main focus of more recent cryptographic research. In the first part of the talk I will survey several results that point out that if we want to prove security under minimal computational assumptions (such as the existence of one-way functions) then inefficiency might be intrinsically unavoidable. These “lower bounds” hold for a large class of constructions and proof techniques. The second part of the talk will survey alternative approaches to achieve efficient and yet provably secure cryptographic algorithms.
Speaker Biography: Rosario Gennaro received his Ph.D. in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology in 1996.
November 2, 2007
Thoughts on the major changes coming in programming practices and computing architectures in that timeframe.
Speaker Biography: Tim Sweeney founded Epic Games in 1991, and wrote a number of early shareware games. In 1995, he began developing the Unreal Engine, which has since grown into the game industry’s leading game technology. Now on its third generation, the Unreal Engine is used in over 150 leading-edge games for PC, Xbox 360, and PlayStation 3; Epic’s current games include Gears of War and Unreal Tournament 3. Nowadays, Sweeney is doing early R&D work on game, graphics, and programming language technology aimed at the large-scale multicore CPUs that will be prevalent in the next decade.
November 15, 2007
November 29, 2007
Although sensor networks are gradually finding their way into commercial products, their applications are still mostly limited to data collection and wire replacement. Part of this limitation is attributed to the lack of systematic frameworks and appropriate sensors that actively interpret and consume sensor data instead of merely transporting it. This talk will describe our current efforts towards building the BehaviorScope, an event-driven network of heterogeneous sensor nodes that tries to observe human behaviors with a precision that is accurate enough to provide meaningful services. Three main components of the BehaviorScope will be presented (sensing, interpretation and middleware) in the context of our driver application, assisted living. We will discuss the requirements, deployment experiences and challenges in data interpretation. With these we will argue that sensing and data interpretation for these types of applications should be well understood before any networking decisions can be made.
Speaker Biography: Andreas Savvides is an Assistant Professor in the Electrical Engineering and Computer Science Departments at Yale University. He is the founder of the Embedded Networks and Applications Lab (ENALAB) that specializes in the design and implementation of distributed sensor networks and smart spaces. Dr. Savvides completed his Ph.D. in the Electrical Engineering Department at UCLA in 2003. Before this he earned his B.S in Computer Engineering from the University of California, San Diego and an M.S in Electrical and Computer Engineering from UMASS, Amherst. Dr. Savvides’ research is supported by an NSF CAREER award as well as other federal grants and industrial support. For 2007, Dr. Savvides is supported by a Junior Faculty Fellowship from Yale University during which he is concentrating on the deployment and application of the BehaviorScope in assisted living applications.