E-commerce has become a crucial part of our life allowing us to buy products, request services, and transfer money easily with a press of a button. As such, establishing immutable trust and reputation of entities that are resilient to manipulation by the malicious are critical in today’s online systems. In this work, we propose a model for calculating trust and reputation using the values stored on blockchain ledger. The model is applied to blockchain-based online payment systems which have a characteristic of immutability by preventing data manipulation. The model normalizes user evaluations based on each user’s personal evaluation criteria that changes over time. In addition, the model derives reputation of, and trust between, users by applying psychological factors. We evaluate our model using not only simulated transaction data but also on real Bitcoin transaction-based dataset to show that our model is able to derive stable values from immutable transactions on blockchain-based online payment systems. Our model has been built into a live commercial blockchain service platform, and new application developments are underway.
Trickle is a scalable and robust flooding algorithm designed for low-power multihop wireless networks. It aims to reduce redundant packet transmissions and collisions while maintaining information consistency, and has been popular due to its simple yet efficient operation on most general topologies. However, Trickle fails to achieve reliability and low latency under certain, but not uncommon, scenarios. To tackle this problem, we propose A2-Trickle, a light-weight enhancement to the Trickle algorithm that guarantees rapid and reliable dissemination under any topology. A2-Trickle aligns the interval boundary at propagation times without synchronizing the global clock, and adapts to the network topology for nodes that could suffer from Trickle’s naïve suppression mechanism. A2-Trickle is implemented on a real embedded device and is evaluated in various scenarios and topologies through both testbed experiments and simulations. The results reveal that A2-Trickle adapts to the network and enables faster and more energyefficient dissemination while maintaining >99% reliability.
17 May 2019 in Journal of Communications and Networks
We investigate the performance of TCP over multihop many-to-one wireless low-power and lossy networks (LLNs), and present TCP assistant-in-the-middle (TAiM) to improve TCP fairness among LLN endpoints without sacrificing aggregate throughput nor end-to-end backward compatibility. TAiM exploits the fact that RTT over LLN is significantly higher, and more variable, than wired or WiFi networks due to unique characteristics of LLN. TAiM intervenes in the middle of TCP communication only at the LLN border router (LBR), and delicately manipulates the RTT of the passing flows to achieve its goal, but does not infringe anything of packet nor operation of existing protocols. TAiM is adaptive and flexible to network status, and does not require any modifications to the TCP/IPv6 stack at the endpoints. We implement TAiM solely in a Linux-based LBR, and perform experiments on two 30-nodes TelosB testbeds running BLIP/TinyRPL stack in TinyOS. Our experiment results show that TAiM helps TCP to operate fairly and efficiently while maintaining total throughput and end-to-end compatibility.
Many IoT applications require a mechanism to disseminate commands and collect responses over a wireless network in order to control and collect data from multiple embedded devices. However, severe collisions may occur if a large number of nodes attempt to respond simultaneously and promptly, not only among the responses, but also with the dissemination of commands. This is because low-power wireless network protocols for dissemination and collection have been designed separately. Tuning the parameters of one side of the protocol has clear trade-off between reliability and latency. To address this challenge, we propose SCoRe, an on-demand scheme for joint scheduling of command and responses on multihop low-power wireless networks to improve both reliability and latency simultaneously at runtime. SCoRe gathers the amount of time required by network nodes for dissemination and collection, and allocates relative timeslots to each node recursively over multihop on-demand when (and only when) disseminating a command. While doing so, information exchange occurs only between local neighbor nodes without a need for global routing table nor time synchronization. We implement SCoRe on a low-power embedded platform, and compare with well-known dissemination and collection schemes through both simulations and testbed experiments on 30 devices. Our evaluation results show that SCoRe can improve both latency and reliability without tuning the parameters for one metric, while the legacy schemes require careful parameter selection to match only one side of SCoRe, never both.
LoRa technology enables low power long range communication thanks to its modulation scheme, the 'Chirp Spread Spectrum (CSS)'. However, it is extremely susceptible and vulnerable to chirp collisions. To resolve this problem, we propose a novel modulation scheme, 'Frequency Hopped Chirp Spread Spectrum (FH-CSS)'. FH-CSS slices a chirp into several sub-chirps and places each to different frequency offsets. This enables demodulated signals of misaligned chirps to spread out in the frequency domain due to the characteristic of CSS. We implement FH-CSS on GNU Radio and compare its performance against standard CSS through real experiments on USRP B200 software defined radios. Preliminary results show that FH-CSS achieves around 90% decoding success rate despite collision while standard CSS has less than 10% on average.