Client Device ANN Model for Privacy-Preserving Ad Targeting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional server-side processing of personal data for mobile advertising raises significant privacy concerns and compliance challenges, especially under stringent regulations like GDPR, as it requires explicit user consent and adequate protections, and remote server modeling performs poorly without access to quality data for detecting complex patterns.
Innovation Solution
A method for performing real-time predictive modeling using a trainable Artificial Neural Network (ANN) model deployed on a client device, which generates inference performance, receives loss calculations, and transmits corrections and coefficients to a remote server, allowing for decentralized data processing while maintaining user privacy and compliance with GDPR.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If server-side processing of personal data is used for mobile advertising, then centralized data processing capability is improved, but privacy concerns and GDPR compliance deteriorate
Solution Approach 1:
The patent implements federated learning where each client device performs local model training using its own personal data, maintaining data locality and privacy while contributing to a global model. The local model updates are aggregated centrally without exposing raw personal data, thus preserving privacy while achieving centralized processing benefits.
Solution Approach 2:
The learning process is segmented into local training phases at client devices and global aggregation phases at the server. This segmentation allows personal data to remain distributed and private at client level while still enabling centralized coordination and model improvement through aggregated learnings.
2Object-affected harmful factors
If remote server modeling is used without access to quality data, then data privacy is improved, but modeling performance deteriorates
Solution Approach 1:
The system implements a feedback loop where the global model is periodically updated with aggregated gradients from multiple client devices, and these updates are sent back to clients for local fine-tuning. This feedback mechanism ensures the global model continuously improves its performance while respecting data privacy constraints.
Solution Approach 2:
The patent introduces gradient updates as an intermediary mechanism that transfers learning information between clients and the server without exposing raw personal data. The server acts as a mediator that aggregates these gradient intermediaries to improve the global model, bridging the gap between privacy preservation and modeling performance.
3Object-affected harmful factors
If decentralized local processing is implemented, then privacy compliance is improved, but device complexity increases
Solution Approach 1:
The system performs preliminary actions by pre-training the global model on aggregated data from multiple clients before deploying it to individual devices. This preliminary training reduces the complexity of local processing required at each device, as the model arrives already equipped with general knowledge that needs only local fine-tuning.
Solution Approach 2:
The patent uses model copying where a standardized global model is replicated across multiple client devices. Each device receives a copy of the global model and performs local training on this copy, avoiding the need to implement complex training infrastructure at each device while maintaining privacy compliance.
Data Source
AI summary
According to an aspect of one or more embodiments, a system for performing a real-time predictive modeling at a client device may include an ANN model in the client device configured to generate an inference performance associated with the trainable ANN model associated with an action performed at the client device based on data comprising a plurality of features associated with the client device, wherein the inference performance is transmitted. The trainable ANN model is configured to receive a loss calculation generated based on the inference performance, indicating a requirement to correct the trainable ANN model, wherein the trainable ANN model generates one or more corrections for coefficients of the trainable ANN model based on the loss calculation. The trainable ANN model is configured to generate one or more coefficients associated with the trainable ANN model and the data, by convoluting the data, and the trainable model. The trainable ANN model is configured to transmit one or more of the one or more corrections and the one or more coefficients indicating a successful predictive modeling of the trainable ANN model with the generation of the trainable ANN model.