ATTENTION: THIS WEB SITE HAS MOVED.
The pages you are looking at are no longer being maintained.
Please go to http://www.poly.edu/cis/
to visit the new site of the Department of Computer and Information Science
at Polytechnic University.
Software Engineering
(Profs. Frankl,
Memon,
Naumovich)
Software engineering research in the department includes work on
software testing techniques, finite state verification techniques, and
network and application security.
Software Testing: Prof. Frankl has worked
extensively on problems in testing large software systems, including
theoretical aspects of software testing, comparing effectiveness of
software testing techniques, and building testing tools. She played an
important part in the development of data flow testing.
In a program, a variable is often defined and used in several different
statements -- data flow testing techniques decide to include a test
case in a test suite if this test case executes certain combinations
of statements for a given variable. Prof. Frankl has also developed a
testing technique for object-oriented programs and worked in the area
of regression testing -- techniques for efficiently testing a new version
of a program when a previous version has already been tested.
Prof. Frankl's and her students' current work includes the development
of techniques for testing database applications, and the
experimental comparison of the effectiveness of several promising
coverage-based testing techniques. Coverage-based techniques measure
the quality of a test set by how fully this set ``exercises'' a
specific aspect of a program. For example, for a test set to achieve
a 100% statement coverage, each statement in the program has
to be executed by at least one test case from this set. The research
is evaluating how good such techniques are at detecting bugs.
One of the
challenges of testing database applications is that the test cases
are represented not only by the conventional parameters passed to
the application, but also by (potentially very large) tables in
the database.
Finite State Verification: This research area addresses the
problem of quality assurance of large distributed software
systems.
Finite state verification (FSV) is a family of techniques that use a
finite model to represent a software system. The finite models are
then analyzed for specification errors. FSV techniques take all
possible inputs of a system into account and are largely automated,
thus providing an important alternative to both testing and formal
verification techniques. Prof. Naumovich has participated in the
development of a FSV technique that can be applied to distributed Ada
and Java programs.
Currently, the main interests of Prof. Naumovich in this area lie in
improving performance of FSV techniques by optimizing algorithms that
they use. Since several studies have demonstrated that even simple
optimizations can dramatically improve performance of FSV tools, he
hopes that a systematic analysis of the nature of different FSV
techniques and previously proposed optimizations will lead to powerful
new optimizations, making FSV applicable to much larger distributed systems
than at present. A substantial part of the work in this direction
will involve extensive experimental work on analyzing realistic
distributed programs with FSV tools.
Another promising research direction is the collaboration of Profs.
Frankl and Naumovich on combining testing and FSV in a synergistic
manner. This approach attempts to use information from FSV analyses to
investigate the reported problems with testing techniques and also to
use information from testing in the form of heuristics to improve
modeling used in FSV. Presently, a tool to be used for empirical
evaluation of this approach is being constructed.
Analysis of Network and Application Security: Profs.
Frankl, Memon, and Naumovich are interested in various aspects
of network and
application security. Currently, their interests are concentrated in
the area of improving confidence in Java programs that may be attacked
by hostile applets. Hostile applets (or, more generally, classes) can
exploit weaknesses in Java systems by obtaining, in one way, or
another, permissions to perform system-critical operations, such as
accessing data on local disks. The main direction of the current work
is to formulate important requirements on Java programs and develop
techniques for validating these requirements. Such techniques will be
based on both finite state verification approaches and testing
approaches.
Virus protection is an important aspect of network security. Prof.
Frankl has worked on detection of macro viruses using a range of
static analysis techniques. Results of this work have been used in a
virus detection tool by IBM.
Software Watermarking: Software piracy is extremely
widespread (estimated at USD 15 Billion in 1999 alone). It is important
that the legitimate developers of a software system be able to
prove their authorship of the software. Software watermarking is
a technique for embedding a message identifying the author in a
software system. Profs. Memon and Naumovich are interested
in the problem of watermarking Java programs. Currently, they are
working on a dynamic watermarking technique, based
on embedding a message in the run-time state of a program.
Profs. Memon and Naumovich are also
interested in several other techniques for preventing software piracy,
including authentication and program obfuscation techniques.
