Richard E. Newman-Wolfe

Implicit locking in the ensemble concurrent object-oriented graphics editor

Ensemble is an X-Windows based, object-oriented graphics editor based on the tgif graphics editor from UCLA. It relies on Unix* 4.3bsd sockets and can be used as a stand-alone program or as an application in the University of Florida’s distributed conferencing system (DCS). It uses implicitly placed write locks for concurrency control, with locks placed when an object is selected and removed when it is deselected. Multiple users may read or edit a file concurrently, with all users receiving updates whenever a lock is removed. Pointers are shared by mutual consent, so that users may collaborate to the degree desired. Ensemble is a prototype lock-based approach to object-oriented concurrent graphics editing.

A protocol for wait-free, atomic, multi-reader shared variables

A wait-free solution to the atomic (r,1kCRWW problem is presented. It uses (r+2)(Sr+2+6)-I safe, multi-reader bits to implement a &bit variable shared among t readers and 1 writer, solving an open question of Lamport [Lamport ‘851. Thus, it closes a gap in the simulation of systems of shared variables for single writers. It differs from other approaches to this problem in that mutual exclusion is maintained between the writer and other processes on the copies of the shared variable used in the implementation.

Performance analysis of a method for high level prevention of traffic analysis using measurements from a campus network

We provide cost estimates for achieving spatial neutrality under realistic network traffic conditions using two methods. Measurements done on the University of Florida campus wide backbone network (UFNET) provide us with considerable experience to model an actual network better. Simulation results show that the algorithms improvement over padding alone is greater for a sparse traffic matrix than for a uniform random traffic matrix. It accomplishes this by smoothing the traffic matrix by rerouting, reducing the padding overhead required to achieve a neutral traffic matrix. On the other hand, a sparse traffic matrix leads to increased costs over uniform random traffic matrix for both padding alone and for padding with rerouting. Experiments done with UFNET traffic characteristics show that the costs are such that the proposed method can be employed in actual networks, under moderate load conditions, to achieve traffic neutrality with acceptable overheads.<<ETX>>

High level prevention of traffic analysis

The paper gives a mathematical model for prevention of traffic analysis in network security and suggests an approach for prevention of unauthorized release of information concerning traffic patterns. The model assumes that an eavesdropper may read the contents of all links, including the source and destination, and that all countermeasures are performed at the transport layer. The goal of the countermeasures is to prevent the eavesdropper from gaining any useful information regarding the traffic patterns in a cost efficient and feasible manner. Countermeasures performed at the transport level include encryption a limited form of message rerouting delaying messages and sending dummy messages as needed within resource capacities. By formulating the problem in terms of systems of equalities and systems of inequalities linear programming methods may be used to find solutions to the traffic analysis security problem.<<ETX>>

Capacity estimation and auditability of network covert channels

Classical covert channel analysis has focused on channels available on a single computer: timing channels and storage channels. We characterize network covert channels. Potential network covert channels are exploited by modulating transmission characteristics. We distinguish between spatial covert channels, caused by a variation in the relative volume of communication between nodes in the network, and temporal covert channels caused by a variation in transmission characteristics over time, extending the work of Girling (1987). A model for obtaining a spatially neutral transmission schedule was given by Newman-Wolfe and Venkatraman (1991, 1992). Temporally neutral transmissions are characterized and scheduling policies to generate temporally neutral transmission schedules were given by Venkatraman and Newman-Wolfe (1993). We estimate the covert channel capacity using an adaptive scheduling policy, modeling the system as a mode secure system. Based on our measurements on the University of Florida campus-wide backbone network (UFNET), we discuss the auditability of network covert channels and suggest some handling policies to reduce the capacity of these covert channels to TCSEC acceptable levels.<<ETX>>

MACE: a fine grained concurrent editor

MACE is a distributed program running on the X Window System and Unix 4.3bsd2 sockets that permits fine-grained (character-level) concurrent editing of text files. It runs both as a stand-alone program and as an application in the University of Florida’s distributed conferencing system (DCS). MACE uses write locks for concurrency control, allowing a locked section of text to be bounded by any pair of characters in the file. Multiple users may read or edit a file concurrently, with all users receiving updates whenever a lock is removed. The level of sharing is controlled by mutual consent, so that users may collaborate e to the degree desired, including the option to view updates in real time. MACE is a first step towards a fine-grained, lock-based approach to concurrent text editing. Permission to copy without fee all or part of this material is that copying is by permission of the Association for Computing granted provided that the copies are not made or distributed for Machinery. To copy otherwise, or to republish, requires a fee direct commercial advantage, the ACM copyright notice and the and/or specific permission. title of the publication and its date appear, and notice is given @ 1991 ACM 0-89791 -456-2 /91/0010 /0240 . ..

Resource allocation in a dynamically partitionable bus network using a graph coloring algorithm

1 .50 ] This work is partially supported by the University of Florida PurdueUniversity Software Engineering Research Center. 2Unix is a trademark of AT&T Bell Labs, 3nemo@chameleon .cis. ufl.edu

Performance analysis of a method for high level prevention of traffic analysis

An efficient dynamic graph traversal algorithm is used to identify nonconflicting requests and to allocate network resources in a dynamically partitionable bus network (DPBN). In centralized network control a special processor receives from the control computer of a partitionable bus network an adjacency matrix which indicates conflicts among requests. It applies the dynamic graph traversal algorithm and returns the identified nonconflicting requests to the control computer. The control computer then physically partitions the network into a number of subnetworks for processing the nonconflicting requests in parallel. In distributed control, each station determines conflicts and sets the switches. The results of performance evaluation show a 40% decrease of network delay as compared with a fully utilized, but unpartitioned local area network. >

An optimal distributed algorithm for failure-driven leader election in bounded-degree networks

One of the stated goals of communications security is the prevention of traffic analysis. A model to prevent traffic analysis by rerouting and padding the traffic matrix, so that the apparent final traffic matrix is neutral, is analyzed. The objective of this analysis is to justify the claims of the model and to show that rerouting of traffic via intermediate nodes with minimum padding is indeed a cost effective method to prevent traffic analysis. Simulation results supporting the above claim are also presented. The requirement that final traffic matrix be neutral may be too restrictive in some cases, and various modifications to the model are suggested, while still ensuring prevention of traffic analysis. Prevention of traffic analysis in real time is addressed briefly.<<ETX>>

A Comparison of Fast and Low Overhead Distributed Priority Locks

The authors consider the leader election problem in point-to-point network where the number of communication links for each node is bounded. Instead of focusing on electing a leader as an initialization step, this paper emphasizes the election of the leader in a fault-tolerant distributed system where the election is triggered by the failure or the abdication of the current leader; it is assumed that the number of initiators that detect the failure or the abdication and start the election algorithm is bounded. A leader election algorithm is presented. The algorithm is based on a new approach that is sharply different from those of previous works: timestamped multiple-sourced flooding. The worst-case time and communication complexities of the algorithm are both optimal; the former is O(D) and the latter is O(E) equivalent to O(N) in the authors network model.<<ETX>>