Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BlocksecRT-DETR: Decentralized Privacy-Preserving and Token-Efficient Federated Transformer Learning for Secure Real-Time Object Detection in ITS

Mohoshin Ara Tahera, Sabbir Rahman, Shuvalaxmi Dass, Sharif Ullah, Mahmoud Abouyessef

TL;DR

This work tackles real-time object detection in ITS under missing-class Non-IID data, edge-latency constraints, and privacy/security risks in decentralized training. It introduces BlockSecRT-DETR, combining RT-DETR with a Token Engineering Module for token-efficient encoding and a blockchain-secured RSU consensus for trust-free aggregation. On a five-client KITTI partition with one missing class per client, it achieves $89.20\%$ mAP@0.5, while TEM reduces encoder FLOPs by $47.8\%$ and latency by $17.2\%$, and the on-chain ledger remains below $12$ KB with around $400$ ms per round. This demonstrates a practical, secure, decentralized approach to high-capacity federated detection in ITS with strong cross-client transfer and scalable security guarantees.

Abstract

Federated real-time object detection using transformers in Intelligent Transportation Systems (ITS) faces three major challenges: (1) missing-class non-IID data heterogeneity from geographically diverse traffic environments, (2) latency constraints on edge hardware for high-capacity transformer models, and (3) privacy and security risks from untrusted client updates and centralized aggregation. We propose BlockSecRT-DETR, a BLOCKchain-SECured Real-Time Object DEtection TRansformer framework for ITS that provides a decentralized, token-efficient, and privacy-preserving federated training solution using RT-DETR transformer, incorporating a blockchain-secured update validation mechanism for trustworthy aggregation. In this framework, challenges (1) and (2) are jointly addressed through a unified client-side design that integrates RT-DETR training with a Token Engineering Module (TEM). TEM prunes low-utility tokens, reducing encoder complexity and latency on edge hardware, while aggregated updates mitigate non-IID data heterogeneity across clients. To address challenge (3), BlockSecRT-DETR incorporates a decentralized blockchain-secured update validation mechanism that enables tamper-proof, privacy-preserving, and trust-free authenticated model aggregation without relying on a central server. We evaluated the proposed framework under a missing-class Non-IID partition of the KITTI dataset and conducted a blockchain case study to quantify security overhead. TEM improves inference latency by 17.2% and reduces encoder FLOPs by 47.8%, while maintaining global detection accuracy (89.20% mAP@0.5). The blockchain integration adds 400 ms per round, and the ledger size remains under 12 KB due to metadata-only on-chain storage.

BlocksecRT-DETR: Decentralized Privacy-Preserving and Token-Efficient Federated Transformer Learning for Secure Real-Time Object Detection in ITS

TL;DR

This work tackles real-time object detection in ITS under missing-class Non-IID data, edge-latency constraints, and privacy/security risks in decentralized training. It introduces BlockSecRT-DETR, combining RT-DETR with a Token Engineering Module for token-efficient encoding and a blockchain-secured RSU consensus for trust-free aggregation. On a five-client KITTI partition with one missing class per client, it achieves mAP@0.5, while TEM reduces encoder FLOPs by and latency by , and the on-chain ledger remains below KB with around ms per round. This demonstrates a practical, secure, decentralized approach to high-capacity federated detection in ITS with strong cross-client transfer and scalable security guarantees.

Abstract

Federated real-time object detection using transformers in Intelligent Transportation Systems (ITS) faces three major challenges: (1) missing-class non-IID data heterogeneity from geographically diverse traffic environments, (2) latency constraints on edge hardware for high-capacity transformer models, and (3) privacy and security risks from untrusted client updates and centralized aggregation. We propose BlockSecRT-DETR, a BLOCKchain-SECured Real-Time Object DEtection TRansformer framework for ITS that provides a decentralized, token-efficient, and privacy-preserving federated training solution using RT-DETR transformer, incorporating a blockchain-secured update validation mechanism for trustworthy aggregation. In this framework, challenges (1) and (2) are jointly addressed through a unified client-side design that integrates RT-DETR training with a Token Engineering Module (TEM). TEM prunes low-utility tokens, reducing encoder complexity and latency on edge hardware, while aggregated updates mitigate non-IID data heterogeneity across clients. To address challenge (3), BlockSecRT-DETR incorporates a decentralized blockchain-secured update validation mechanism that enables tamper-proof, privacy-preserving, and trust-free authenticated model aggregation without relying on a central server. We evaluated the proposed framework under a missing-class Non-IID partition of the KITTI dataset and conducted a blockchain case study to quantify security overhead. TEM improves inference latency by 17.2% and reduces encoder FLOPs by 47.8%, while maintaining global detection accuracy (89.20% mAP@0.5). The blockchain integration adds 400 ms per round, and the ledger size remains under 12 KB due to metadata-only on-chain storage.
Paper Structure (31 sections, 8 equations, 2 figures, 4 tables, 1 algorithm)

This paper contains 31 sections, 8 equations, 2 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works and Challenges
  3. Transformer-Based Object Detection and Efficiency Techniques
  4. Federated Learning for Vision Tasks: Challenges and Security
  5. System Overview and Threat Model
  6. BlockSecRT-DETR Framework Overview
  7. System Entities and Roles
  8. System Workflow
  9. Threat Model and Security Assumptions
  10. Threats and attack mitigation
  11. Security Assumptions
  12. Methodology
  13. Phase I: System Initialization
  14. Phase II: FL Client Training
  15. Phase III: Privacy-Preserving Update Generation
  16. ...and 16 more sections

Figures (2)

  • Figure 1: Overview of BlockSecRT-DETR: Phase I (executed once): System Initialization. Phases II–V (executed iteratively in each federated learning round): (II) FL Client Model Training using RT-DETR model integrated with TEM, (III) Privacy-Preserving Update Generation, (IV) Verification and Aggregation, and (V) RSU Consensus and Model Finalization.
  • Figure 2: Block generation time versus the number of clients ($N$) with a fixed RSU committee size of $K=3$ and $R=15$ training rounds.