Partha Dasgupta

Topology of the conceptual network of language

We define two words in a language to be connected if they express similar concepts. The network of connections among the many thousands of words that make up a language is important not only for the study of the structure and evolution of languages, but also for cognitive science. We study this issue quantitatively, by mapping out the conceptual network of the English language, with the connections being defined by the entries in a Thesaurus dictionary. We find that this network presents a small-world structure, with an amazingly small average shortest path, and appears to exhibit an asymptotic scale-free feature with algebraic connectivity distribution.

The Clouds distributed operating system

The authors discuss a paradigm for structuring distributed operating systems, the potential and implications this paradigm has for users, and research directions for the future. They describe Clouds, a general-purpose operating system for distributed environments. It is based on an object-thread model adapted from object-oriented programming.<<ETX>>

Connectivity distribution and attack tolerance of general networks with both preferential and random attachments

A general class of growing networks is constructed with both preferential and random attachments, which includes random and scale-free networks as limiting cases when a physical parameter is tuned. Formulas are derived characterizing the evolution and distribution of the connectivity, which are verified by numerical computations. Study of the effect of random failures and intentional attacks on the performance of network suggests that general networks which are neither completely random nor scale-free are desirable.

Structuring Peer-to-Peer Networks Using Interest-Based Communities

Interest-based communities are a natural arrangement of distributed systems that prune the search space and allow for better dissemination of information to participating peers. In this paper, we introduce the notion of peer communities. Communities are like interest groups, modeled after human communities and can overlap. Our work focuses on providing efficient formation, discovery and management techniques that can be implemented to constantly changing community structures. We provide a mechanism to generate realistic peer-to-peer network topologies that can be used in simulations that evaluate the operation of our algorithms. Our experiments show how searching the peer-to-peer network can take advantage of peer communities to reduce the number of messages and improve the quality of search results.

On using reputations in ad hoc networks to counter malicious nodes

Nodes in mobile ad hoc networks have a limited transmission range. Hence the nodes expect their neighbors to relay packets meant for far off destinations. These networks are based on the fundamental assumption that if a node promises to relay a packet, it relays it and does not cheat. This assumption becomes invalid when the nodes in the network have tangential or contradictory goals. The reputations of the nodes, based on their past history of relaying packets, can be used by their neighbors to ensure that the packet is relayed by the node. This paper introduces a reputation scheme for ad hoc networks. Instead of choosing the shortest path to the destination, the source node chooses a path whose next hop node has the highest reputation. This policy, when used recursively, in the presence of 40% malicious nodes, improves the throughput of the system to 65%, from 22 % throughput provided by AODV. This improvement is obtained at the cost of a higher number of route discoveries with a minimal increase in the average hop length according S. Bansal and M. Baker (2003).

Parallel processing on networks of workstations: a fault-tolerant, high performance approach

One of the most sought after software innovation of this decade is the construction of systems using off-the-shelf-workstations that actually deliver and even surpass, the power and reliability of supercomputers. Using completely novel techniques: eager scheduling, evasive memory layouts and dispersed data management it is possible to build an execution environment for parallel programs on workstation networks. These techniques were originally developed in a theoretical framework for an abstract machine which models a shared memory asynchronous multiprocessor. The network of workstations platform presents an inherently asynchronous environment for the execution of our parallel program. This gives rise to substantial problems of correctness of the computation and of proper automatic load balancing of the work amongst the processors, so that a slow processor will not hold up the total computation. A limiting case of asynchrony is when a processor becomes infinitely slow, i.e. fails. Our methodology copes with all these problems, as well as with memory failures. An interesting feature of this system is that it is neither a fault-tolerant system extended for parallel processing nor is it parallel processing system extended for fault tolerance. The same novel mechanisms ensure both properties.

P2P Reputation Management Using Distributed Identities and Decentralized Recommendation Chains

Peer-to-peer (P2P) networks are vulnerable to peers who cheat, propagate malicious code, leech on the network, or simply do not cooperate. The traditional security techniques developed for the centralized distributed systems like client-server networks are insufficient for P2P networks by the virtue of their centralized nature. The absence of a central authority in a P2P network poses unique challenges for reputation management in the network. These challenges include identity management of the peers, secure reputation data management, Sybil attacks, and above all, availability of reputation data. In this paper, we present a cryptographic protocol for ensuring secure and timely availability of the reputation data of a peer to other peers at extremely low costs. The past behavior of the peer is encapsulated in its digital reputation, and is subsequently used to predict its future actions. As a result, a peers reputation motivates it to cooperate and desist from malicious activities. The cryptographic protocol is coupled with self-certification and cryptographic mechanisms for identity management and countering Sybil attack. We illustrate the security and the efficiency of the system analytically and by means of simulations in a completely decentralized Gnutella-like P2P network.

Process migration: a generalized approach using a virtualizing operating system

Process migration has been used to perform specialized tasks, such as load sharing and checkpoint/restarting long running applications. Implementation typically consists of modifications to existing applications and the creation of specialized support systems, which limit the applicability of the methodology. Off the shelf applications have not benefited from process migration technologies, mainly due to the lack of an effective generalized methodology and facility. The benefits of process migration include mobility, checkpointing, relocation, scheduling and on the fly maintenance. This paper shows how regular shrink-wrapped applications can be migrated. The approach to migration is to virtualize the application by injecting functionality into running applications and operating systems. Using this scheme, we separate the physical resource bindings of the application and replace it with virtual bindings. This technique is referred to as virtualization. We have developed a virtualizing operating system (vOS), residing on top of Windows 2000 that injects stock applications with the virtualizing software. It coordinates activities across multiple platforms providing new functionality to the existing applications. The vOS makes it possible to build communities of systems that cooperate to run applications and share resources non-intrusively while retaining application binary compatibility.

A role-based trust model for peer-to-peer communities and dynamic coalitions

Although P2P systems are usually used for information exchange between peers, they have either protected peers anonymity, or required transacting peers to trust each other implicitly. Both these approaches are vulnerable to attacks by malicious peers who could abuse the P2P system to spread viruses, incorrect, or damaging information. Here, we propose an approach for trust management in P2P systems. We introduce an optimistic role-based model for trust amongst peers and show that it is scalable, dynamic, revocable, secure and transitive. Our proposed solution permits asymmetric trust relationships that can be verified by any peer in the system through a simple, low-cost algorithm. We introduce a metric known as iComplex that combines a peers trust value for each of its roles into a single, relative, probabilistic guarantee of trust. Finally, we discuss how our trust model allows peers to revoke relationships with malicious peers, and the nonrepudiation of peer relations. We use simulations to illustrate the trust value distribution amongst peers in the network. Our analysis and experiments demonstrates the low-cost involved to verify and validate trust values. Lastly, we establish the effectiveness of using sum as the aggregation function to combine trust values of a peer.

Securing P2P networks using peer reputations: is there a silver bullet?

Peer reputations have been used as security tools not only to motivate peers against cheating but also to protect good peers from the chronic cheaters. Although the reputation management techniques are not confined to P2P networks, they present novel challenges that were absent in central server based distributed systems. We enumerate these challenges and survey the solutions proposed by the community to counter them. These challenges include, but are not limited to, peer-identification in decentralized environments, reputation metrics, storage and exchange of reputation data. Finally we survey the applications which use P2P network paradigm and therefore can benefit from the reputation systems.