DNS Request Probability Estimation for Ad Blocking

The patent changes the parameter of DNS request handling by introducing probability estimation and weighting factors. Instead of blanket blocking all DNS requests from advertising modules, the system estimates the probability that each DNS request is user-initiated versus advertising-initiated, and applies selective blocking based on these probabilities. This parameter-based approach reduces unnecessary blocking of user requests while still effectively blocking advertising requests, thereby maintaining blocking effectiveness while reducing power consumption compared to universal blocking methods.

The patent applies local quality by treating different DNS requests differently based on their characteristics. The system evaluates each DNS request individually using weighting factors that consider the source application, domain type, and request patterns. User-initiated DNS requests receive different treatment compared to advertising module DNS requests. This localized, differentiated approach ensures that only advertising-related data transmissions are blocked while user data transmissions proceed normally, reducing overall data transmission costs while maintaining blocking effectiveness.

The patent implements feedback mechanisms by continuously monitoring DNS request patterns, application behaviors, and user interactions. The weighting factors are dynamically adjusted based on feedback from observed request patterns. When a DNS request is made, the system evaluates multiple factors including the source application's historical behavior, the domain being accessed, and the timing of the request. This feedback-driven approach allows the system to learn and distinguish between genuine user requests and advertising requests, improving protection against intrusive ads while minimizing incorrect blocking of user requests.

The patent applies partial action by implementing a tiered probability estimation approach. Instead of performing exhaustive analysis on every DNS request, the system first applies quick heuristic checks using pre-defined weighting factors for common scenarios. Only requests that pass initial filtering or exhibit unusual patterns undergo more complex probability estimation. This partial application of comprehensive analysis maintains high accuracy in distinguishing user requests from ad requests while avoiding the computational overhead of full analysis on every request, thus managing device complexity effectively.