A Smart Contract Vulnerability Detection Model Based on Syntactic and Semantic Fusion Learning

Table 3

Syntax representation module experimental results.

	Methods	Accuracy	Precision	Recall	F1
ImplicitVisibility	LSTM-AST	85.50	84.82	88.06	86.41
	GRU-AST	86.24	85.32	86.46	86.95
	Our model	89.00	87.58	92.04	89.75
IntegerOverflow	LSTM-AST	94.17	93.52	95.58	94.54
	GRU-AST	93.92	92.28	95.05	94.16
	Our model	95.58	94.08	97.79	95.90
IntegerUnderflow	LSTM-AST	96.17	92.76	87.23	89.91
	GRU-AST	95.83	89.70	88.94	89.32
	Our model	96.42	92.86	88.51	90.63
TimeDependency	LSTM-AST	94.29	85.48	88.79	87.10
	GRU-AST	95.92	89.61	89.22	89.42
	Our model	98.25	97.52	94.85	94.81
Reentrancy	LSTM-AST	97.83	68.97	54.05	60.61
	GRU-AST	97.92	63.64	70.27	69.14
	Our model	98.64	90.91	78.38	75.36