Michael Langberg

Resilient network coding in the presence of Byzantine adversaries

Network coding substantially increases network throughput. But since it involves mixing of information inside the network, a single corrupted packet generated by a malicious node can end up contaminating all the information reaching a destination, preventing decoding. This paper introduces the first distributed polynomial-time rate-optimal network codes that work in the presence of Byzantine nodes. We present algorithms that target adversaries with different attacking capabilities. When the adversary can eavesdrop on all links and jam zO links , our first algorithm achieves a rate of C - 2zO, where C is the network capacity. In contrast, when the adversary has limited snooping capabilities, we provide algorithms that achieve the higher rate of C - zO. Our algorithms attain the optimal rate given the strength of the adversary. They are information-theoretically secure. They operate in a distributed manner, assume no knowledge of the topology, and can be designed and implemented in polynomial-time. Furthermore, only the source and destination need to be modified; non-malicious nodes inside the network are oblivious to the presence of adversaries and implement a classical distributed network code. Finally, our algorithms work over wired and wireless networks.

Resilient Network Coding in the Presence of Byzantine Adversaries

Network coding substantially increases network throughput. But since it involves mixing of information inside the network, a single corrupted packet generated by a malicious node can end up contaminating all the information reaching a destination, preventing decoding. This paper introduces the first distributed polynomial-time rate-optimal network codes that work in the presence of Byzantine nodes. We present algorithms that target adversaries with different attacking capabilities. When the adversary can eavesdrop on all links and jam zO links , our first algorithm achieves a rate of C - 2zO, where C is the network capacity. In contrast, when the adversary has limited snooping capabilities, we provide algorithms that achieve the higher rate of C - zO. Our algorithms attain the optimal rate given the strength of the adversary. They are information-theoretically secure. They operate in a distributed manner, assume no knowledge of the topology, and can be designed and implemented in polynomial-time. Furthermore, only the source and destination need to be modified; non-malicious nodes inside the network are oblivious to the presence of adversaries and implement a classical distributed network code. Finally, our algorithms work over wired and wireless networks.

Approximation Algorithms for Maximization Problems Arising in Graph Partitioning

Given a graph G=(V,E), a weight function w: E?R+, and a parameter k, we consider the problem of finding a subset U?V of size k that maximizes:Max-Vertex Coverk: the weight of edges incident with vertices in U,Max-Dense Subgraphk: the weight of edges in the subgraph induced by U,Max-Cutk: the weight of edges cut by the partition (U,V\U),Max-Uncutk: the weight of edges not cut by the partition (U,V\U).For each of the above problems we present approximation algorithms based on semidefinite programming and obtain approximation ratios better than those previously published. In particular we show that if a graph has a vertex cover of size k, then one can select in polynomial time a set of k vertices that covers over 80% of the edges.

The encoding complexity of network coding

In the multicast network coding problem, a source s needs to deliver h packets to a set of k terminals over an underlying communication network G. The nodes of the multicast network can be broadly categorized into two groups. The first group includes encoding nodes, i.e., nodes that generate new packets by combining data received from two or more incoming links. The second group includes forwarding nodes that can only duplicate and forward the incoming packets. Encoding nodes are, in general, more expensive due to the need to equip them with encoding capabilities. In addition, encoding nodes incur delay and increase the overall complexity of the network. Accordingly, in this paper, we study the design of multicast coding networks with a limited number of encoding nodes. We prove that in a directed acyclic coding network, the number of encoding nodes required to achieve the capacity of the network is bounded by h/sup 3/k/sup 2/. Namely, we present (efficiently constructible) network codes that achieve capacity in which the total number of encoding nodes is independent of the size of the network and is bounded by h/sup 3/k/sup 2/. We show that the number of encoding nodes may depend both on h and k by presenting acyclic coding networks that require /spl Omega/(h/sup 2/k) encoding nodes. In the general case of coding networks with cycles, we show that the number of encoding nodes is limited by the size of the minimum feedback link set, i.e., the minimum number of links that must be removed from the network in order to eliminate cycles. We prove that the number of encoding nodes is bounded by (2B+1)h/sup 3/k/sup 2/, where B is the minimum size of a feedback link set. Finally, we observe that determining or even crudely approximating the minimum number of required encoding nodes is an /spl Nscr/P-hard problem.

On the hardness of approximating the network coding capacity

This work addresses the computational complexity of achieving the capacity of a general network coding instance. It has been shown [Lehman and Lehman, SODA 2005] that determining the “scalar linear” capacity of a general network coding instance is NP-hard. In this paper we address the notion of approximation in the context of both linear and nonlinear network coding. Loosely speaking, we show that given an instance of the general network coding problem of capacity C , constructing a code of rate αC for any universal (i.e., independent of the size of the instance) constant α ≤ 1 is “hard”. Specifically, finding such network codes would solve a long standing open problem in the field of graph coloring. Our results refer to scalar linear, vector linear, and nonlinear encoding functions and are the first results that address the computational complexity of achieving the network coding capacity in both the vector linear and general network coding scenarios. In addition, we consider the problem of determining the (scalar) linear capacity of a planar network coding instance (i.e., an instance in which the underlying graph is planar). We show that even for planar networks this problem remains NP-hard.

An equivalence between network coding and index coding

We show that the network coding and index coding problems are equivalent. This equivalence holds in the general setting which includes linear and nonlinear codes. Specifically, we present a reduction that maps a network coding instance to an index coding instance while preserving feasibility, i.e., the network coding instance has a feasible solution if and only if the corresponding index coding instance is feasible. In addition, we show that one can determine the capacity region of a given network coding instance with colocated sources by studying the capacity region of a corresponding index coding instance. Previous connections between network and index coding were restricted to the linear case.

Local graph coloring and index coding

We present a novel upper bound for the optimal index coding rate. Our bound uses a graph theoretic quantity called the local chromatic number. We show how a good local coloring can be used to create a good index code. The local coloring is used as an alignment guide to assign index coding vectors from a general position MDS code. We further show that a natural LP relaxation yields an even stronger index code. Our bounds provably outperform the state of the art on index coding but at most by a constant factor.

Correction of adversarial errors in networks

We design codes to transmit information over a network, some subset of which is controlled by a malicious adversary. The computationally unbounded, hidden adversary knows the message to be transmitted, and can observe and change information over the part of the network being controlled. The network nodes do not share resources such as shared randomness or a private key. We first consider a unicast problem in a network with |epsiv parallel, unit-capacity, directed edges. The rate-region has two parts. If the adversary controls a fraction p < 0.5 of the |epsiv edges, the maximal throughput equals (1 - p) |epsiv|. We describe low-complexity codes that achieve this rate-region. We then extend these results to investigate more general multicast problems in directed, acyclic networks

On the complementary Index Coding problem

The Index Coding problem is one of the basic problems in wireless network coding. In this problem, a server needs to deliver a set P of packets to several clients through a noiseless broadcast channel. Each client needs to obtain a certain subset of P and has prior side information about a different subset of P. The objective is to satisfy the requirements of all clients with the minimum number of transmissions. Recently, it was shown that the Index Coding problem is NP-hard. Furthermore, this problem was shown to be hard to approximate under a widely accepted complexity assumption.

The RPR 2 rounding technique for semidefinite programs

Several combinatorial optimization problems can be approximated using algorithms based on semidefinite programming. In many of these algorithms a semidefinite relaxation of the underlying problem is solved yielding an optimal vector configuration v1,...,vn. This vector configuration is then rounded into a {0, 1} solution. We present a procedure called RPR2 (Random Projection followed by Randomized Rounding) for rounding the solution of such semidefinite programs. We show that the random hyperplane rounding technique introduced by Goemans and Williamson, and its variant that involves outward rotation, are both special cases of RPR2. We illustrate the use of RPR2 by presenting two applications. For Max-Bisection we improve the approximation ratio. For Max-Cut, we improve the tradeoff curve (presented by Zwick) that relates the approximation ratio to the size of the maximum cut in a graph.