Research Projects

In today’s digital ecosystem, platforms such as e-commerce marketplaces, social media, and content aggregators heavily rely on user-generated trust signals like ratings, reviews, and credibility scores to drive content visibility, recommendations, and decision-making. However, these signals are often siloed across platforms, vulnerable to manipulation, and subject to systemic biases, especially against underrepresented groups. Furthermore, centralizing trust data raises serious concerns about user privacy and platform dominance. This project introduces Federated Social Trust Aggregation (FSTA), a novel framework that synthesizes federated learning, social choice theory, and privacy-preserving computation to securely and fairly aggregate trust signals across multiple decentralized platforms. FSTA addresses a critical and timely challenge: How can we compute robust, fair, and privacy-aware reputation scores for socially relevant entities (e.g., news articles, sellers, or services) without centralizing sensitive user data? The project has three main scientific objectives: 1. To build a federated model that allows platforms to learn and contribute trust scores without revealing user-level data. 2. To design voting-based aggregation mechanisms that incorporate user and platform credibility while satisfying fairness principles (e.g., proportionality, anonymity, core-stability). 3. To analyze trade-offs between fairness, privacy, and robustness under adversarial behaviors, such as sybil attacks or strategic manipulation. The central hypothesis is that fair and privacy-preserving trust aggregation is possible across platforms if social choice-based aggregation is combined with federated modeling and differential privacy techniques. The research follows a three-stage methodology: 1. Local Modeling: Each platform learns a local model to estimate user trust scores while keeping raw data private. 2. Fair Aggregation: Aggregated scores are computed using fairness-aware voting schemes, weighted by reputation. 3. Privacy & Auditability: Secure computation or differential privacy mechanisms are used to prevent identity or score leakage, while blockchain-inspired logs ensure transparency. The successful project will advance the scientific understanding of federated trust modeling and enable practical applications in news credibility, platform governance, and decentralized rating systems. It aligns closely with national priorities around digital trust, AI fairness, and privacy-by-design, and has potential to inform policy and strengthen ethical AI ecosystems.