Title: Quantum-Safe IoT with Federated
Learning and TinyML for Detecting HDNL-like
Anomalies
Authors: Sheikh Sayed Bin Rahman, Azizbek Reimbaev, Kanita Jerin Tanha, Minsoo Kim, Taesoo Jun
Presented By: Sheikh Sayed Bin Rahman
z z
Contents
1. Introduction
2. Motivations & Contributions
3. Proposed System Model
4. Dataset Description
5. Results and analysis
6. Future Works
7. Conclusion
z z
2
Introduction
Internet of Vehicles expansion enables real-time telemetry,
moving insurance from fixed premiums to dynamic
behaviour-based models that reward safer driving and
increase fairness across policyholders.
Driving Behaviour Analyze Behaviour
Traditional centralised behaviour-based insurance
suffers from single-point failures, data manipulation
risks, opacity in premium calculation, and driver
distrust, limiting adoption and fairness.
BBI Solutions
Vehicles (IoV Sensors) Data Protection BBI Bidding
Decentralised ledgers offer immutable trip records,
transparent automated contracts, and tamper-resistant
claims, aligning incentives and restoring trust
between insurers and vehicle owners
z z
3
Motivation & Contribution
Motivations Contributions
Centralized BBI systems are vulnerable to The study proposes a secure, blockchain-enabled BBI
manipulation and suffer from low trust due framework using PureChain and its Proof-of-Authority and
to opaque decision-making. Association (PoA²) consensus.
There is a pressing need for a transparent, The solution integrates IoV sensors, a secure off-chain
fault-tolerant solution that preserves data aggregator, a permissioned PureChain blockchain, and
integrity and privacy in real-time IoV smart contracts for dynamic policy and premium
environments. management.
The latency and throughput limitations of The system effectively balances scalability, immutability,
public blockchains hinder real-time privacy, and transparency, demonstrated through real
insurance applications in vehicular vehicular telemetry data experiments.
networks.
z z
4
Proposed system model
• Vehicle telemetry (e.g., speed, acceleration, accident
events, heart rate) is collected by IoV sensors and
processed off-chain for privacy.
• Behaviour scores are computed and summarized data is
stored on-chain via PureChain, which ensures rapid
validation (PoA²), scalability, and immutability.
• Smart contracts automate insurance logic, enabling
dynamic adjustment of premiums and automated claims
processing.
• The architecture features an insurer dashboard and
payment gateway for efficient management.
Fig. 1. PureChain-Enabled Behavior-Based Insurance System
Architecture for IoV.
z z
5
Proposed system model
Data Flow & Consensus Path
Telemetry packets stream to aggregator for local scoring;
anonymised summaries are timestamped and validated by
PoA² nodes, then stored immutably to support contract
queries and audits.
Smart Contract Modules
Separate contracts handle policy storage, behaviour scoring,
premium logic, claim verification and payment release,
enabling modular upgrades while preserving deterministic
execution.
Fig. 1. PureChain-Enabled Behavior-Based Insurance System
Architecture for IoV.
z z
6
Dataset Description
Purpose: It is designed for applications like analyzing traffic safety, developing predictive models for accident risk, and understanding
the relationship between different factors and the likelihood of a crash.
Aspect Description
Duration and Size 18 hours of driving data with over 61,000 observations and 32 attributes.
Integrates driver information (age, gender, medical conditions, violations), vehicle details (brand, model, type, year), weather conditions, traffic
Data Sources
accidents, and road geometry.
Collection Method Data acquired using an OBD-II scanner, GPS receiver, and health monitor through a custom acquisition module.
Collection Location Recorded on two high-traffic, high-accident routes in Quito, Ecuador.
Labeled Data 1,980 labeled samples categorized into four risk levels: low, medium, high, and very high.
Dataset: POLIDriving: A Public-Access Driving Dataset for Road Traffic Safety Analysis
z z
7
Results and analysis
The final risk score was scaled to a 0–100 range for
interpretability:
Table 1: Sampled Driver Data With Computed Risk Scores
Weighted aggregation R=0.5·accidents+0.3·heartRate+0.2·speed yields
0-100 risk index; scores below 50 indicate safe drivers, above 75 trigger
high-premium and intervention alerts.
z z
8
Results and analysis Cont.
Table 2 : Classification Performance of RF Per Class
z z
9
Results and analysis Cont.
The table-3 comparing the performance
metrics of the PureChain and Sepolia
blockchain networks, showing that
PureChain uses less gas, costs less per
transaction, has lower execution cost, much
lower latency, and significantly higher
throughput than Sepolia
Table 3: Performance Comparison Betwen PureChain and Sepolia
The Figure-2 visualizes a threat score
for behavior-based insurance
assessment, indicating a low risk score
of 98.5 out of 100, with itemized counts
for various risk and impact categories,
most of which fall under "no impact"
and "beneficial," and very few in any
risk category.
Fig. 2. Threat Score Visualization for Behavior-Based Insurance Assessment
z z
10
Conclusion
PureChain BBI framework delivers trustworthy, low-latency, high-throughput behaviour-based vehicle insurance,
automating fair premiums and tamper-proof claims while incentivising safer roads.
z z
11
Future Works
Privacy-Preserving Enhancements
Integrate zero-knowledge proofs and federated learning to hide raw driving data while retaining verifiable
scoring, meeting GDPR and personal data protection mandates.
Personalised Insurance Products
Expand contracts to support usage-based, peer-to-peer and on-demand coverage, leveraging rich telematics for
contextual pricing that reflects individual driving patterns and environment.
z z
12
감사합니다