Communication, Control and Signal Processing Seminar
Abstracts of talks
02/09 Himanshu Tyagi (UMD), Fault-tolerant secret key generation
Mobile nodes observing correlated data communicate using an insecure
bidirectional switch to generate a secret key, which remains concealed from the switch. We are interested in fault-tolerant
secret key rates, i.e., the rates of secret key generated even if a subset of nodes drop out before
the completion of the communication protocol. We formulate a new notion of fault-tolerant secret key capacity, and present an upper bound on it. This
upper bound is shown to be tight when the random variables corresponding to the observations of nodes are exchangeable.
Further, it is shown that one round of interaction achieves the fault-tolerant secret key capacity in this case. The
upper bound is also tight for the case of a pairwise
independent network model
consisting of a complete graph, and can be attained
by a noninteractive protocol.
This is joint work with Navin Kashyap (IISc Bangalore), Yogesh Sankarasubramaniam (HP Labs, India), Kapali Viswanathan (HP Labs, India).
02/23 Alexander Barg (UMD)
Polar codes for q-ary channels, q=2^r
(UMD) Polar codes for q-ary channels, q=2^r
Constructing polar codes for binary-input channels can be
accomplished in a number of ways related to the choice of the polarizing kernel.
At the same time, for nonbinary channels the capacity-achieving performance is
shown by using averaging arguments to prove the existence of polarizing kernels
with the needed properties. In this work we propose an explicit scheme for
attaining channel capacity with polar codes on q-ary channels, q=2^r.
Joint work with Woomyoung Park.
3/1/12 Raef Bassily, A cryptographic
treatment of the wiretap channel: An optimal, explicit, and efficient encryption
Abstract: In this talk, I will present the recent work of M. Bellare, S.Tessaro, and A. Vardy where they establish important results that bridge some gaps between the information-theoretic and cryptographic perspectives to the security problem in the context of wiretap channels. I will also present a subsequent work of Bellare and Tessaro where they provide the first explicit polynomial-time construction of an encryption scheme that attains the secrecy capacity of a class of wiretap channels under a security definition that is shown to be stronger than the standard one in the information-theoretic security literature.
In particular, the following contributions are made:
First, new, strong definitions of security in the wiretap model are suggested. One, called MIS (mutual information security), is an extension of a weaker definition called MIS-R (mutual information security for random messages) which is the one in current use in the information theory community. Another, called SS (semantic security), adapts a definition of security that is well-understood in the cryptography community, vetted by decades of cryptographic research and targeted by standards deployed in practice. Next, it is shown that MIS and SS definitions are indeed equivalent.
3/8/12 Vinay A. Vaishampayan (AT&T Labs), Query Matrices for the Hamming Oracle
x is an unknown binary string of length n. In response
to an n-bit query y, the Hamming oracle returns the Hamming distance d(x,y). The
problem is to construct a set of queries that allow us to determine x based on
the query results. All queries must be set up in advance, i.e. queries are not
allowed to depend on the response of the oracle to preceding queries. Our figure
of merit is the query ratio, i.e. the ratio of the number of queries to the
length of the string.
Connections to the distinct-subset-sum problem and to other more widely known problems will be explored. Bounds on the achievable query ratio as a function of the string length n will be derived. An algebraic construction will be presented along with a decoding algorithm, and it will be shown that query ratios arbitrarily close to zero can be achieved.
3/15/12 Raef Bassily, cont'd
3/22/12 Spring break
3/29/12 Yuval Lomnitz (Tel Aviv University), Universal communication over unknown channels (with feedback)
Abstract: A communication problem over an unknown channel is discussed. Two main settings are presented. In the "individual channel" setting, there is no model for the channel, and information rates are defined as a function of the channel input sequence and channel output sequence,
i.e. "what really happened". The range of rate-functions, defining the rate as a function of the inputs and outputs is explored. In the "unknown vector channel" setting, the channel has a vector-wise probabilistic model, defining the probability of every output sequence given every input sequence (not necessarily stationary), however this probability is unknown to the transmitter and the receiver. Using low rate feedback, and common randomness, the goal is to approach the same rates that would be obtained by someone knowing the channel, but constrained in encoding and decoding complexity.
4/5, 4/12 David Ward (UMD) Introduction to Team Decision Problems
Abstract: In this introduction to team decision problems, static team decision problems will be discussed and it will be proven that under
quadratic costs the Bayes team decision function is obtained as a projection. Under the further assumptions of jointly Gaussian distributions
on the information variables and additional assumptions on the cost function, the Bayes team decision function will be shown to be linear in the
information variables. This discussion will follow Team Decision Problems by Radner.
Also, partially nested information patterns will be discussed. It will be shown that under the assumption of a linear information pattern, partially
nested information patterns are the only ones equivalent to static information patterns.
4/19, 4/26 Maya Kabkab (UMD) Equilibrium
selection in repeated coordination games
Abstract: In this talk I will discuss equilibrium selection in repeated coordination games. I will begin by formally defining what a pure
coordination game means, and provide examples. We will then allow a coordination game, with binary choices, to be played repeatedly by
identical players. We will be interested in the long-run behavior of the fraction of players playing each action. The equilibrium selection
criteria will be formally characterized for the different dynamics we will consider. The discussion will follow Equilibrium Selection in n-Person Coordination Games by Youngse Kim.
5/3 Anup Menon (UMD) Cheeger Inequality and it's implications in the study of Expander Families of Graphs
Abstract: We motivate the study of expanders with a couple of examples relevant to us. We define a combinatorial quantity called edge expansion of a graph and give the combinatorial definition of expander families. The main result that we discuss is Cheeger's inequality. It is a relationship between edge expansion and spectral properties of the graph adjacency matrix for regular graphs and leads to a spectral characterisation of expander families. We follow Luca Trevisan's lecture notes 3 and 4: http://theory.stanford.edu/~trevisan/cs359g/