Multi-run security
Paper in proceedings, 2011

This paper explores information-flow control for batch-job programs that are allowed to be re-run with new input provided by the attacker. We argue that directly adapting two major security definitions for batch-job programs, termination-sensitive and termination-insensitive noninterference, to multi-run execution would result in extremes. While the former readily scales up to multiple runs, its enforcement is typically over-restrictive. The latter suffers from insecurity: secrets can be leaked in their entirety by multiple runs of programs that are secure according to batch-job termination-insensitive noninterference. Seeking to avoid the extremes, we present a framework for specifying and enforcing multi-run security in an imperative language. The policy framework is based on tracking the attacker's knowledge about secrets obtained by multiple program runs. Inspired by previous work on robustness, the key ingredient of our type-based enforcement for multi-run security is preventing the dangerous combination of attacker-controlled data and secret data from affecting program termination. © 2011 Springer-Verlag.

Author

Arnar Birgisson

Chalmers, Computer Science and Engineering (Chalmers), Software Technology (Chalmers)

Andrei Sabelfeld

Chalmers, Computer Science and Engineering (Chalmers), Software Technology (Chalmers)

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

03029743 (ISSN) 16113349 (eISSN)

372-391

Areas of Advance

Information and Communication Technology

Subject Categories

Computer and Information Science

Roots

Basic sciences

DOI

10.1007/978-3-642-23822-2_21

ISBN

978-364223821-5

More information

Created

10/7/2017