Donghang Duan

Current LLM-based frameworks for text anonymization usually rely on remote API services from powerful LLMs, which creates an inherent privacy paradox: users must disclose the raw data to untrusted third parties for guaranteed privacy preservation. Moreover, directly migrating current solutions to local small-scale models (LSMs) offers a suboptimal solution with severe utility collapse. Our work argues that this failure stems not merely from the capability deficits of LSMs, but significantly from the inherent irrationality of the greedy adversarial strategies employed by current state-of-the-art (SOTA) methods. To address this, we propose Rational Localized Adversarial Anonymization (RLAA), a fully localized and training-free framework featuring an Attacker-Arbitrator-Anonymizer architecture. We model the anonymization process as a tradeoff between Marginal Privacy Gain (MPG) and Marginal Utility Cost (MUC), demonstrating that greedy strategies tend to drift into an irrational state. Instead, RLAA introduces an arbitrator that acts as a rationality gatekeeper, validating the attacker’s inference to filter out ghost leaks. This mechanism promotes a rational early-stopping criterion, and structurally prevents utility collapse. Extensive experiments on different benchmarks demonstrate that RLAA achieves a superior privacy-utility trade-off compared to strong baselines.

Blockchain-based Federated Learning (BFL) facilitates collaborative model training with guaranteed transparency and accountability across distributed participants. However, BFL encounters a critical security vulnerability: advanced free-riders can exploit the public on-chain model histories to synthesize fabricated gradients without performing actual local training, which is indistinguishable from legitimate contributions. Previous efforts to defense free-rider attack mostly focus on anomaly detection, but they struggle to detect some advanced free-riders. To address this fundamental challenge, we propose BFLGR (Blockchain-based Federated Learning with Gresham’s Reversal), which is a novel framework integrating reverse auction mechanisms and a dynamic reputation system. BFLGR employs a dynamic selection algorithm that can eliminate advanced free-riders by exploiting their profit-seeking motives and parasitic dependency on honest participants’ contributions. This framework ensures genuine contributions prevail over deceptive submissions, and achieves a paradigmatic shift in BFL security, transforming the transparency-exploitation vulnerability into a strategic advantage for detecting and eliminating malicious participants. Theoretical proofs and extensive experiments show that BFLGR performs well in advanced free-rider attack situation and outperforms our baselines.