2026-10-06T18:38:08Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34030518 2026-10-01T06:10:54Z category_29203 item_type_3 month_year_10_2026

BinFuncGT N A (25137249) Software engineering not elsewhere classified code and data <pre><table><tr><th><pre><table><tr><th><table><tr><th>BinFuncGTBinFuncGT is a research framework for binary software composition analysis (BSCA). It provides function-level ground truth, static feature extraction, candidate retrieval, call-graph-aware structural matching, and evaluation of component and version identification. The project, benchmark, and construction utilities are named BinFuncGT. Python modules and filesystem paths use the lowercase spelling binfuncgt. FeaturesBinary analysis: Parse ELF metadata, symbol tables, and DWARF information, and recover functions and call graphs with angr CFGFast.Function representation: Extract instruction counts, basic-block counts, control-flow edges, call sites, immediate-constant statistics, cyclomatic complexity, and mnemonic histograms.Candidate retrieval: Generate candidate function pairs using cosine similarity, a top-r selection rule, and a similarity threshold.Structural matching: Combine local similarity, preserved call relationships, and one-to-one matching constraints in a unified optimization objective.Reduction and classical solvers: Apply safe reductions and interaction-graph decomposition, with greedy, Hungarian, CP-SAT, simulated annealing, and small-instance exhaustive solvers.Function-level evaluation: Measure precision, recall, and F1 against raw and strict ground-truth correspondences.Version identification: Explore constant features, string features, differential weighting, and version-discriminative string scoring.Statistical analysis: Provide paired comparisons, ranking diagnostics, and Benjamini–Hochberg false discovery rate correction.Repository StructureThe main directories and modules are organized as follows: .├── configs/ # Benchmark and experiment configurations├── data/ # Binary data, intermediate features, and instances├── experiments/│ ├── pilot/ # Matching, baseline, ablation, and ranking results│ │ └── score_cache_v2/ # Per-instance scoring caches│ ├── case_study/ # Debian case-study results and reports│ └── versiongrid_v3/ # Extended version-grid design and build records├── results/ # Aggregate results, statistics, and logs├── scripts/ # Experiment and report-generation entry points├── src/binfuncgt/│ ├── binary/ # ELF parsing, call graphs, and ground truth│ ├── data/ # BinFuncGT schemas and instance I/O│ ├── features/ # Static features and vectorization│ ├── retrieval/ # Candidate function retrieval│ ├── matching/ # Matching instances and objectives│ ├── reduction/ # Safe and heuristic reductions│ ├── solvers/ # Classical optimization methods│ ├── scoring/ # Downstream structural scores│ ├── evaluation/ # Evaluation metrics│ └── feasibility/ # Classical matching and solver statistics└── requirements.txt # Pinned Python dependenciesConfigurationThe principal benchmark configurations are: ConfigurationExperiment settingconfigs/binkit_x86_v2.yamlSame-architecture matching across build variantsconfigs/binkit_xarch_v2.yamlCross-architecture matching with the same compiler and optimization levelconfigs/pilot_versiongrid_v2.yamlComponent and version comparisons on VersionGrid v2configs/pilot_versiongrid_v3.yamlExtended version-grid comparisonsThe x86 configuration name is historical: its pairing rule is arch_relation: same, covering matching architectures represented in the input corpus. Common settings include: ParameterMain protocol settingCandidate count per target functionr = 5Cosine-similarity threshold0.5Same-architecture feature setfullCross-architecture feature setstructuralMatching objective weightsalpha = beta = 1CP-SAT time budget60 seconds by default in the main scoring runnersStructural score weights(0.5, 0.3, 0.2)Experiment ScriptsScriptPurposescripts/rq1_function_gt.pyFunction-level matching evaluationscripts/rq1_function_gt_report.pyFunction-level result summaries and paired comparisonsscripts/run_feasibility.pyStructural matching and per-instance feasibility statisticsscripts/analyze_results.pyPilot result aggregationscripts/score_v2.pyMulti-method benchmark scoringscripts/report_v2.pyScoring-cache summaries and statistical comparisonsscripts/run_baselines.pyClassical baselines through run and summarize subcommandsscripts/rq5_feature_attribution.pyFeature attribution analysisscripts/rq6_stripped.pyEvaluation with stripped target binariesscripts/rq7_diff_weighted.pyDifferential feature-weighting experimentsscripts/rq7_tie_audit.pyTie and candidate-set diagnosticsscripts/rq7_str_scorer.pyVersion-discriminative string scoringscripts/rq7_stats.pyStatistical comparisons of string-based methodsscripts/fdr_correction.pyBenjamini–Hochberg FDR correctionscripts/debian_case_study.pyDebian case-study evaluationscripts/debian_str_idf.pyString-IDF evaluation on Debian binaries</th></tr></table></th></tr></table></pre></th></tr></table></pre> 2026-10-01T06:10:54Z Dataset Dataset 10.6084/m9.figshare.34030518.v3 https://figshare.com/articles/dataset/BinFuncGT/34030518 CC BY 4.0