Blockchain-Linked Auditable Decision Management for Telecom/IoT Fraud-Control Requests
Proposes a blockchain-linked, auditable request management framework integrating ML, LLM, and blockchain for telecom/IoT fraud control, enabling request-level policy and traceability.
Key Findings
Methodology
The framework maps each synthetic deployment record to a managed request, applying a deterministic hard-fraud gate to block out-of-boundary cases. Non-hard requests are scored via three risk sources: M1 (centralized ML), M2 (federated meta-learning), and M3 (LLM-based). Based on scores, requests are assigned to one of five states, with two-zone refinement for ambiguous cases. The entire process is governed by a unified algorithm (Algorithm 1). Blockchain records off-chain decisions, ensuring traceability. Experiments on synthetic training and deployment replay data show M1 balances best in validation, with FPR 0.0890 and recall 0.8341, but FPR increases in deployment; M3-QLoRA improves soft-fraud recall to 0.8240 while maintaining low FPR. Blockchain telemetry indicates lifecycle costs are driven by off-chain profiles, not fraud logic.
Key Results
- Validation results show M1 achieves FPR 0.0890 and recall 0.8341, but in deployment FPR rises to 0.1646, while M3-QLoRA reaches 0.1801 FPR with a soft-fraud recall of 0.8240, demonstrating effective risk score fusion. The blockchain telemetry confirms that lifecycle metrics depend more on off-chain decision profiles than on fraud logic changes. The multi-source risk scoring and blockchain audit integration significantly enhance transparency and traceability in fraud management.
Significance
This work shifts fraud control from detector-centric to request-level policy management, integrating multiple risk sources with blockchain-based auditability. It addresses the industry’s need for transparent, traceable, and adaptable fraud mitigation in future 6G/IoT networks. By enabling request lifecycle management and auditability, it enhances system trustworthiness, accountability, and operational efficiency. The approach offers a scalable, modular framework for deploying multi-source risk assessments in real-world telecom and IoT environments, paving the way for more intelligent and trustworthy network security systems.
Technical Contribution
The paper introduces a multi-risk source fusion mechanism combining centralized ML, federated meta-learning, and LLM-based risk scoring, integrated into a unified request management workflow. It designs a deterministic hard-fraud gate and a five-state decision policy with two-zone refinement, enabling fine-grained request control. The blockchain audit layer records each request’s lifecycle, ensuring transparency and traceability. The experimental validation on synthetic data demonstrates the framework’s ability to balance detection accuracy, operational transparency, and auditability, offering a novel architecture for trustworthy telecom/IoT fraud management.
Novelty
This is the first comprehensive framework to fuse heterogeneous risk signals—centralized ML, federated meta-learning, and LLMs—within a request-level, blockchain-audited fraud control workflow. It introduces a deterministic hard-fraud gate combined with a five-state, two-zone refinement policy, enabling precise, transparent decision-making. Unlike prior works focusing solely on detection, this approach emphasizes lifecycle management and auditability, providing a scalable, modular system adaptable to future 6G/IoT environments.
Limitations
- The validation is conducted solely on synthetic data and deployment replay environments, lacking real-world operational validation, which may limit generalizability. The system’s complexity and parameter tuning requirements pose deployment challenges. Blockchain interactions introduce additional latency and costs, which need optimization for large-scale deployment.
Future Work
Future efforts will focus on deploying and validating the framework in real operational telecom and IoT networks, optimizing blockchain interactions for efficiency, and integrating richer, multi-modal risk signals. Further research will explore adaptive threshold tuning, real-time processing, and robustness against adversarial attacks, aiming to realize a fully trustworthy, scalable, and intelligent fraud management system.
AI Executive Summary
The rapid evolution of 6G and IoT technologies has transformed communication networks into highly dense, low-latency, and security-critical environments. Traditional fraud detection methods, primarily detector-centric, struggle to meet the demands for request-level policy enforcement, lifecycle traceability, and comprehensive auditability. This gap hampers effective management of evolving fraud tactics and regulatory compliance. Addressing this challenge, the authors propose a novel blockchain-linked, auditable request management framework that integrates multiple risk sources—centralized machine learning, federated meta-learning, and large language models (LLMs)—to evaluate each telecom/IoT service record during deployment.
The core innovation lies in mapping each service record to a managed request, applying a deterministic hard-fraud gate to block overt violations, and scoring non-hard requests through diverse risk sources. These scores inform a five-state policy, with two-zone refinement for ambiguous cases, enabling nuanced decision-making. The entire process is governed by a unified algorithm (Algorithm 1), ensuring consistency and transparency. Blockchain technology underpins the system by recording each request’s lifecycle, decision logs, and audit trail, making the entire process auditable and tamper-proof.
Experimental validation on synthetic datasets and deployment replay data demonstrates the framework’s effectiveness. In validation, the centralized ML configuration (M1) achieves a low false positive rate (FPR 0.0890) and high recall (0.8341). However, in deployment scenarios, FPR increases, with M3-QLoRA significantly improving soft-fraud recall to 0.8240 while maintaining acceptable FPR. These results highlight the importance of multi-source risk fusion and blockchain-based auditability for trustworthy fraud management.
This research paves the way for more transparent, accountable, and adaptive telecom and IoT security systems. By transforming fraud control into a request-level, lifecycle-aware process, it addresses long-standing industry challenges. Future work will focus on real-world deployment, optimizing blockchain interactions, and incorporating richer risk signals to enhance robustness and scalability, ultimately contributing to the development of intelligent, trustworthy communication networks.
Deep Dive
Key Concepts
硬性欺诈门控 (Hard-Fraud Gate)
一种基于规则的机制，直接阻断明显的欺诈请求，确保系统安全性。它在请求管理中起到第一道防线，避免误判风险。
五状态策略 (Five-State Policy)
一种请求决策模型，将请求划分为五个不同状态（如批准、阻断、可能低风险等），实现细粒度管理。
两区细化机制 (Two-Zone Refinement)
对模糊或不确定的请求，采用两级细化策略，提升决策的准确性和透明度。
区块链审计层 (Blockchain Audit Layer)
利用区块链技术记录请求的生命周期和决策过程，确保每一步都可追溯、不可篡改。
多风险源融合 (Multi-Source Risk Fusion)
结合中心化ML、联邦学习和LLM等多种风险评估机制，提升欺诈识别的全面性和准确性。
Open Questions Unanswered questions from this research
- 1 当前模型主要在合成环境验证，缺乏真实运营数据验证，实际效果仍需进一步验证。
- 2 多源风险信号融合的参数调节复杂，系统部署时面临调优难题。
- 3 区块链交互带来的延迟和成本问题尚未优化，影响大规模应用的可行性。
Applications
Immediate Applications
电信运营商欺诈监控
实时评估和管理通信中的欺诈请求，提升检测透明度和追溯能力。
物联网安全管理
对物联网设备的通信请求进行生命周期管理，确保安全和合规。
Long-term Vision
可信网络生态构建
打造全链路可追溯、可信的通信基础设施，支持自动化审计和合规。
Abstract
Telecom fraud-control studies often stop at detector-level classification, but deployment use requires request-level policy resolution, lifecycle traceability, and auditability. This paper reframes fraud control as blockchain-linked auditable decision management for synthetic telecom/IoT fraud-control requests, and its main result is that the QLoRA-tuned LLM branch becomes much more usable than zero-shot prompting but mainly approaches, rather than outperforms, a lower-cost centralized ensemble. The framework maps each synthetic deployment record to a managed request, blocks explicit out-of-boundary cases through a deterministic hard-fraud gate, scores non-hard requests using centralized ML (M1), federated meta-learning (M2), or LLM-family risk sources (M3), and resolves actions through a shared five-state policy, two-zone refinement mechanism, and local Ethereum-compatible audit layer. Evaluation uses separate synthetic training data and a 100,000-record deployment replay corpus, so the study should be read as controlled drift-replay evidence rather than field validation or proof of live deployability. On validation, M1 gives the strongest balance, with legitimate-request FPR 0.0890 under the 0.10 operating cap and soft-fraud recall 0.8341. On labeled deployment replay, however, the legitimate-FPR gap becomes large: M1 rises to 0.1646 and M3-QLoRA to 0.1801, while M3-QLoRA reduces the M3-Base legitimate FPR from 0.3915 and reaches 0.8240 soft-fraud recall. Blockchain telemetry shows that lifecycle gas, cost, latency, and throughput differences are driven by submitted off-chain decision profiles rather than changes in fraud logic.