Indirect Transfer Association AI Algorithm in Dynamic Translation System
Yan Hu · 2025
In dynamic binary translation systems, indirect transmission often leads to reduced execution efficiency and insufficient accuracy. This study adopts a method based on Long Short Term Memory (LSTM) to optimize the prediction accuracy and conversion efficiency of indirect transmission. Firstly, collect a large number of indirect jump instruction sequences from various operating environments. Secondly, through data cleaning and other processing techniques, the data quality and model generalization ability have been improved. Finally, a model architecture was designed that includes multiple layers of LSTM units. The experimental results show that the LSTM model is significantly superior to traditional RNN (Recurrent Neural Network) and GRU (Gated Recurrent Unit) models on datasets of different scales. In large-scale data processing, the prediction accuracy, translation delay time, and cache hit rate of the LSTM model are 91.0%, 8000ms, and 85%, respectively. In the above data conclusions, LSTM provides an effective solution that significantly improves the accuracy of indirect jump prediction and the overall performance of dynamic binary translation systems.