Figure feature bidirectional fusion embedding-based firmware vulnerability similarity detection method

The invention relates to a firmware vulnerability similarity detection method based on graph feature bidirectional fusion embedding, and the method specifically comprises the steps: extracting a firmware function control flow graph: carrying out the disassembly of a firmware function through a related tool, namely IDA Pro, extracting the control flow graph, and carrying out the instruction feature conversion of each statement, performing instruction statement feature vector conversion: performing context influence extraction on instruction statement vectors through an attention mechanism, performing basic block feature fusion by mixing with LSTM, and performing graph embedding feature vector extraction: performing vector fusion on firmware function features by using a graph embedding network; and similarity detection is carried out through the Siamese network to judge whether known vulnerabilities exist or not. According to the invention, the vulnerability similarity detection precision can be improved. 本发明涉及