Dhawaj Kumar Tomar
Research Subjects
Dhawaj Kumar Tomar has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Ai Liu
College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics
2 shared publications - Shaoying Liu
East China Normal University
2 shared publications Xi Hu
1 shared publicationYang Liu
1 shared publication- Yang Liu
College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu, China
1 shared publication Zedong Jia
1 shared publication
Latest Publications
- Open Access
- 5 August 2026
- Published by Elsevier in SSRN Electronic Journal
FSF-Guided Program Slicing for Testing-Based Formal Verification of Functional Soundness and Completeness
Soundness and completeness are two sides of the same coin, used in program analysis to balance between catching all errors and avoiding false positives. Testing-based formal verification has been proposed to reduce testing costs and improve software reliability by automatically verifying the correctness of program execution paths with regard to the functional scenario form of a specification. In this paper, we propose the notions of functional soundness and functional completeness with regard to a functional scenario and develop the corresponding analysis techniques by integrating testing-based formal verification with program slicing. To further reduce the program scale and verification cost, we propose an FSF-guided program slicing technique, which generates a scenario-specific executable slice for each functional scenario $(T_i,D_i)$. The technique identifies the input and output variables involved in the functional scenario, preserves their relevant data and control dependencies, and removes branches that are unreachable under the testing condition $T_i$.Moreover, we implement tool support for the Java language and conduct an experiment on 250 Java programs containing 740 scenario-specific slicing tasks. The experiment investigates three research questions concerning the reduction in program scale, the reduction in verification cost, and the preservation of verification results. The experimental results show that FSF-guided slicing reduces the numbers of lines of code, executable statements, and cyclomatic complexity by 39.21\%, 46.81\%, and 45.81\% on average, respectively. It also reduces the total cost of testing-based formal verification by 22.5\%, achieves a 97.03\% conclusive preservation rate over all tasks, and preserves the functional soundness and functional completeness judgments on all definite verification tasks.- 5 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessFSF-Guided Program Slicing for Testing-Based Formal Verification of Functional Soundness and Completeness
Soundness and completeness are two sides of the same coin, used in program analysis to balance between catching all errors and avoiding false positives. Testing-based formal verification has been proposed to reduce testing costs and improve software reliability by automatically verifying the correctness of program execution paths with regard to the functional scenario form of a specification. In this paper, we propose the notions of functional soundness and functional completeness with regard to a functional scenario and develop the corresponding analysis techniques by integrating testing-based formal verification with program slicing. To further reduce the program scale and verification cost, we propose an FSF-guided program slicing technique, which generates a scenario-specific executable slice for each functional scenario $(T_i,D_i)$. The technique identifies the input and output variables involved in the functional scenario, preserves their relevant data and control dependencies, and removes branches that are unreachable under the testing condition $T_i$.Moreover, we implement tool support for the Java language and conduct an experiment on 250 Java programs containing 740 scenario-specific slicing tasks. The experiment investigates three research questions concerning the reduction in program scale, the reduction in verification cost, and the preservation of verification results. The experimental results show that FSF-guided slicing reduces the numbers of lines of code, executable statements, and cyclomatic complexity by 39.21\%, 46.81\%, and 45.81\% on average, respectively. It also reduces the total cost of testing-based formal verification by 22.5\%, achieves a 97.03\% conclusive preservation rate over all tasks, and preserves the functional soundness and functional completeness judgments on all definite verification tasks. - Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
A CCT-Centered Framework for Testing-Based Formal Verification: Bounded Exploration, Parallel Test Generation, and Fault Localization
Testing-based formal verification (TBFV) combines concrete execution with formal checking of represented program paths. Condition Sequence Coverage (CSC) and the Condition Case Tree (CCT) make this exploration systematic while controlling loop-induced growth. We present a CCT-centered framework that extends CSC-based TBFV in three ways. CSC+Boundary reconsiders ancestor-stopped obligations under a selected finite input constraint and classifies them as concrete, infeasible over the specification input domain, or range-excluded. Parallel CCT-frontier generation executes snapshot-visible candidates concurrently and merges their traces deterministically. Fold-the-Tree fault localization folds refined-TBFV outcomes, CCT topology, and assignment/return trace events into ranked condition and state-update candidates. We implement the framework in Java and evaluate it on 42 subjects and 252 retained mutants, 225 of which provide failure evidence for diagnosis. Bounded exploration leaves loop-free subjects unchanged and resolves all observed ancestor-stopped obligations in loop-bearing subjects. Across 210 loop-bearing mutants, CSC+Boundary increases the kill rate from 77.6% to 92.4%. With eight workers, parallel generation achieves mean and median speedups of 1.80× and 1.89×, while all canonical output fingerprints match across worker settings and repetitions. Under a matched source-line budget, the CCT-based composite has higher observed Top-1 and Top-2 hit rates than the strongest evaluated spectrum-based fault localization (SBFL) baseline and ties it at Top-3; after Holm correction, none of the paired differences is significant at the 0.05 level.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA CCT-Centered Framework for Testing-Based Formal Verification: Bounded Exploration, Parallel Test Generation, and Fault Localization
Testing-based formal verification (TBFV) combines concrete execution with formal checking of represented program paths. Condition Sequence Coverage (CSC) and the Condition Case Tree (CCT) make this exploration systematic while controlling loop-induced growth. We present a CCT-centered framework that extends CSC-based TBFV in three ways. CSC+Boundary reconsiders ancestor-stopped obligations under a selected finite input constraint and classifies them as concrete, infeasible over the specification input domain, or range-excluded. Parallel CCT-frontier generation executes snapshot-visible candidates concurrently and merges their traces deterministically. Fold-the-Tree fault localization folds refined-TBFV outcomes, CCT topology, and assignment/return trace events into ranked condition and state-update candidates. We implement the framework in Java and evaluate it on 42 subjects and 252 retained mutants, 225 of which provide failure evidence for diagnosis. Bounded exploration leaves loop-free subjects unchanged and resolves all observed ancestor-stopped obligations in loop-bearing subjects. Across 210 loop-bearing mutants, CSC+Boundary increases the kill rate from 77.6% to 92.4%. With eight workers, parallel generation achieves mean and median speedups of 1.80× and 1.89×, while all canonical output fingerprints match across worker settings and repetitions. Under a matched source-line budget, the CCT-based composite has higher observed Top-1 and Top-2 hit rates than the strongest evaluated spectrum-based fault localization (SBFL) baseline and ties it at Top-3; after Holm correction, none of the paired differences is significant at the 0.05 level.