Bounded Model Construction for Monadic Second-Order Logics

Abstract

The monadic logics M2L-Str and WS1S have been successfully used for verification, although they are nonelementary decidable. Motivated by ideas from bounded model checking, we investigate procedures for bounded model construction for these logics. The problem is, given a formula φ and a bound k, does there exist a word model for φ of length k. We give a bounded model construction algorithm for M2L-Str that runs in a time exponential in k. For WS1S, we prove a negative result: bounded model construction is as hard as validity checking, i.e., it is nonelementary. From this, negative results for other monadic logics, such as S1S, follow. We present too preliminary tests using a SAT-based implementation of bounded model construction; for certain problem classes it can find counterexamples substantially faster than automata-based decision procedures.

QBF Reasoning on Real-World Instances

“…To test the effectiveness of the techniques implemented in QUBE++ we used the set of 450 formal verification and planning instances that constituted part of the QBF evaluation 1 : 25% of these instances are from formal verification problems [3,4], and the remaining are from planning domains [5,6]. All the experiments where run on a farm of PCs, each one equipped with a PIV 2.4GHz processor, 1GB of RAM, and running Linux RedHat 7.2.…”

“…Following the standard practice of SAT competitions [1,2], we consider as real-world the instances originated by translations from application domains such as, e.g., Formal Verification [3,4], Planning [5,6], and Reasoning about Knowledge [7]. In the last evaluation of QBF solvers [8], instances of this kind emerged as challenging benchmarks for the current state-of-the-art tools: real-world benchmarks represented about 50% of the evaluation test set, and they constituted about 95% of the "hard" instances, i.e., problems that could not be solved by any of the participants within the allotted time.…”

Algorithmic Verification of Intransitive Noninterference for 3-domain Security Policies with a SAT Solver

“…Our work is motivated by which the verification methods based on SAT solvers [15][16][17][18] have been shown to push the envelope of functional verification in terms of both capacity and efficiency, as reported in several academic and industrial case studies [19][20][21][22]. The successful application of SAT solvers in formal verification is due to dramatic improvements in SAT solver technology over the past decade.…”

A Hierarchy of Tractable Subclasses for SAT and Counting SAT Problems

“…It has applications in many areas of computer science, such as program verification using bounded model checking [22,25] and bounded model construction [6,7], real-time and embedded systems verification [3,4,23], planning problem in artificial intelligence [3,19], and so on.…”