Abstract
By utilizing a robust packet exchange mechanism, the Novel Two-way Time Synchronization Protocol (NTSP) makes time synchronization more precise and steady in Wireless Sensor Networks (WSNs). NTSP adopts a simulated annealing (SA) algorithm to defend against random noise and malicious delay attacks. Network simulations demonstrate that NTSP achieves high synchronization accuracy, rapid convergence, and strong defensive capabilities against malicious delays, outperforming traditional protocols.
Introduction
In Wireless Sensor Networks (WSNs), precise timing is critical for data fusion—the process of combining data from multiple sensors based on the exact moment events occurred. However, cheap hardware leads to clock drift, where oscillators run at slightly different speeds, and environmental factors like temperature cause these offsets to fluctuate. The Novel Two-way Time Synchronization Protocol (NTSP) addresses these challenges by combining a two-way packet exchange mechanism with a continuous clock model and a Simulated Annealing (SA) optimization algorithm.
The Synchronization Challenge
Traditional protocols often struggle with specific trade-offs. For instance, the Timing-sync Protocol for Sensor Networks (TPSN) uses two-way communication but ignores clock skew (speed differences), leading to errors between sync cycles. Conversely, the Flooding Time Synchronization Protocol (FTSP) models clock skew but relies on one-way broadcasting, which can leave nodes "blind" if a parent node fails. Furthermore, most protocols are vulnerable to malicious nodes that inject artificial delays to disrupt network timing.
As shown in the figure above, synchronization must account for multiple sources of delay, including send, access, transmission, propagation, and reception delays. NTSP is designed to calculate and mitigate these factors effectively.
The NTSP Methodology
NTSP introduces a hybrid approach that balances complexity and precision through four main steps:
- Hierarchical Topology: The network organizes itself into layers starting from a "root" node (Layer 0).
- Two-Way Exchange: Unlike one-way broadcasts, NTSP uses a request-response mechanism. This ensures nodes can detect if their parent is still active and allows for the calculation of both propagation delay and clock offset.
- Clock Modeling: NTSP estimates the relative clock skew (the rate of change), allowing a node to predict the parent's time even between synchronization pulses.
- Simulated Annealing (SA) Optimization: This is the protocol's core innovation. NTSP uses an influence factor ($\alpha$) to weight current clock data against historical data. The SA algorithm dynamically tunes this factor to minimize synchronization error, allowing the system to ignore "outlier" data caused by network congestion or malicious attacks.
The clock model, illustrated above, accounts for the transition from the physical quartz crystal oscillator to the quantized time value used by the node software.
Evaluation and Results
The authors evaluated NTSP using the NS2 (Network Simulator 2) in a 5x5 grid of 25 nodes. To simulate real-world conditions, nodes were assigned random clock skews (20 ppm standard deviation), and malicious delays were introduced using Gaussian distributions.
Key results include:
- Multi-Hop Stability: In a 4-hop network, NTSP maintained significantly lower and more stable synchronization errors than both TPSN and FTSP.
- Rapid Convergence: NTSP reaches a synchronized state faster than FTSP because it avoids long "root-timeout" waiting periods.
- Resilience to Attacks: During simulated malicious delay injections, the SA-enhanced version of NTSP showed much smaller error peaks and faster recovery times. The SA-tuned influence factor effectively "smoothed out" the impact of the attacks.
Discussion and Future Work
NTSP successfully addresses the "blindness" problem of one-way protocols by using a two-way mechanism that allows nodes to proactively seek new parents if communication fails. While the protocol demonstrates superior accuracy and robustness, future research is needed to analyze the specific computational and energy costs of running the Simulated Annealing algorithm on resource-constrained microcontrollers. Additionally, the authors plan to explore extending synchronization periods to further reduce power consumption without sacrificing precision.
Summary updated