FedCRM-DP: A Privacy-Preserving Federated Learning Framework for Secure Cross-Organization CRM Intelligence
Abstract
Abstract Collaborative intelligence across regulated enterprises remains constrained by stringent data-sharing restrictions imposed by frameworks such as GDPR, HIPAA, and CCPA. This paper presents FedCRM-DP, a privacy-preserving federated learning framework designed for secure cross-organization Customer Relationship Management (CRM) systems. The framework unifies three complementary mechanisms: FedProx-based optimization for non-IID enterprise data heterogeneity, calibrated Differential Privacy via gradient clipping and Gaussian noise injection, and Secure Aggregation (simulated via pairwise random masking) to prevent server-side inspection of individual model updates. Together, these components enable institutions—banks, hospitals, insurers, and large-scale platform operators—to jointly train predictive CRM models without exposing raw customer records. We validate the framework on the UCI Bank Marketing dataset, modeled as a lead conversion prediction task distributed across five demographically partitioned enterprise clients. Experiments compare six learning strategies using Accuracy, F1-score, AUC-ROC, convergence efficiency, and privacy-utility trade-off metrics. FedCRM-DP achieves Accuracy 0.8850, F1-score 0.5720, and AUC-ROC 0.8760 at privacy budget ε = 5—closely approaching the centralized upper bound while preserving meaningful privacy guarantees that centralized deployment cannot provide. Ablation results confirm that ε = 5 represents the Pareto-optimal operating…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.