Table 8 McNemmar Test (Accuracy Difference between Optimised-CodeBERT and Optimised-LSTM).
From: Deep learning-based solution for smart contract vulnerabilities detection
Vulnerability type | Discordant pairs (B) | Continuity correction (C) | Chi-squared value (\({\mathscr {X}}^2\)) | p-value | Accuracy difference |
|---|---|---|---|---|---|
Re-entrancy | 2638 | 1 | 1761.44 | \(\approx 0\) | Extremely significant |
Timestamp-dependency | 147 | 1 | 285.33 | \(\approx 0\) | Extremely significant |
tx.origin | 1 | 1 | 0 | \(\approx 1\) | Not significant |
Unhandled-exceptions | 205 | 1 | 203.25 | \(\approx 0\) | Extremely significant |
