Advertisement Plug-in Recognition via Multi-Dimensional Feature Vectors

The patent segments the advertisement plug-in into multiple feature dimensions including title module, icon, package name, class name, and behavior characteristics. Each dimension is analyzed separately using feature vectors, allowing comprehensive identification without relying on a single detection method. This segmentation enables the system to identify obfuscated advertisements by examining multiple aspects simultaneously.

The patent transitions from one-dimensional title detection to multi-dimensional feature analysis by introducing feature vectors across multiple dimensions (title, icon, package name, class name, behavior). This dimensional expansion allows the system to capture complex advertisement characteristics that cannot be detected by simple title matching alone, thereby improving recognition accuracy while managing complexity through structured analysis.

The patent performs preliminary extraction and organization of feature vectors from plug-in files before conducting similarity calculations. By pre-processing and structuring the feature data (titles, icons, package names, class names, behaviors) into standardized vectors, the system reduces the computational burden during the actual recognition phase, thereby decreasing processing time while maintaining high accuracy.

The patent transforms complex plug-in data into standardized feature vectors with specific parameters and dimensions. By changing the representation form of advertisement data from raw code to structured feature vectors, the system enables efficient comparison and similarity calculation, reducing computational complexity while preserving recognition accuracy.

The patent implements a hierarchical recognition approach where not all feature dimensions are analyzed with equal depth for every plug-in. The system performs preliminary screening on basic features (title, package name) and only conducts comprehensive multi-dimensional analysis on suspicious cases. This partial action strategy maintains high reliability for identifying malicious software while reducing overall computational resource consumption by avoiding exhaustive analysis of all plug-ins.

The patent introduces feature vectors as an intermediary representation between raw plug-in code and final recognition results. These feature vectors serve as a compressed, structured intermediary that captures essential characteristics without requiring full code analysis. This intermediary layer reduces computational resources needed while maintaining reliable identification capability through efficient similarity comparison.