Webpage Spoofing Detection via Multi-Dimensional Feature Fingerprinting

The patent segments the webpage detection process into multiple independent feature extraction components: text features, markup features, visual features, and redirect path features. Each feature type is extracted and analyzed separately, then combined to form a comprehensive fingerprint for classification. This segmentation allows the system to achieve high detection accuracy by examining multiple dimensions of the webpage without requiring a single overly complex detection mechanism.

The patent transitions from traditional single-dimension URL-based detection to multi-dimensional webpage analysis. By extracting and analyzing text, markup, visual, and redirect path features simultaneously, the system creates a multi-dimensional fingerprint space where spoofed webpages can be distinguished from legitimate ones. This dimensional expansion enables detection of advanced spoofing techniques that evade traditional URL blacklists.

The patent performs preliminary feature extraction during the webpage access phase, capturing text, markup, visual, and redirect path features before the classification decision is made. By preparing these features in advance and storing them in a structured fingerprint format, the system reduces the computational burden during the actual classification process. The features are extracted once and reused for comparison against the corpus of known webpages.

The patent creates simplified copies of webpage features in the form of fingerprints that can be rapidly compared against a corpus of known webpages. Instead of analyzing entire webpages during classification, the system uses compact feature representations (fingerprints) that capture the essential characteristics. This copying approach enables efficient comparison and classification while maintaining high detection accuracy.