Search

Abstract This paper identifies and analyzes applications in which Large Language Models (LLMs) can make Internet of Things (IoT) networks more intelligent and responsive through three case studies from critical topics: DDoS attack detection, macroprogramming over IoT systems, and sensor data processing. Our results reveal that the GPT model under few-shot learning achieves 87.6% detection accuracy, whereas the fine-tuned GPT increases the value to 94.9%. Given a macroprogramming framework, the GPT model is capable of writing scripts using high-level functions from the framework to handle possible incidents. Moreover, the GPT model shows efficacy in processing a vast amount of sensor data by offering fast and high-quality responses, which comprise expected results and summarized insights. Overall, the model demonstrates its potential to power a natural language interface. We hope that researchers will find these case studies inspiring to develop further.

In decentralized finance (DeFi), stablecoins like DAI are designed to offer a stable value amidst the fluctuating nature of cryptocurrencies. We examine the class of crypto-backed stable derivatives, focusing on mechanisms for price stabilization and exemplified by the well-known stablecoin DAI from MakerDAO. For simplicity, we consider a single-collateral setting. We introduce a belief parameter to the simulation model of DAI in a previous work (DAISIM), reflecting market sentiments about the value and stability of DAI, and show that it better matches the expected behavior when this parameter is set within a particular range of values. Our methods include comparing simulated data with real-world data, focusing on monthly correlations between ETH and DAI prices and scatter plots illustrating the relationship of their price trends over time. We also propose a simple mathematical model of DAI price to explain its stability and dependency on ETH price. Finally, we analyze possible risk factors associated with these stable derivatives to provide valuable insights for stakeholders in the DeFi ecosystem.

Researchers have been interested in developing AI tools to help students learn various mathematical subjects. One challenging set of tasks for school students is learning to solve math word problems. We explore how recent advances in natural language processing, specifically the rise of powerful transformer based models, can be applied to help math learners with such problems. Concretely, we evaluate the use of GPT-3, GPT-3.5, and GPT-4, all transformer models with billions of parameters recently released by OpenAI, for three related challenges pertaining to math word problems corresponding to systems of two linear equations. The three challenges are classifying word problems, extracting equations from word problems, and generating word problems. For the first challenge, we define a set of problem classes and find that GPT models generally result in classifying word problems with an overall accuracy around 70%. There is one class that all models struggle about, namely the “item and property” class, which significantly lowered the value. For the second challenge, our findings align with researchers’ expectation: newer models are better at extracting equations from word problems. The highest accuracy we get from fine-tuning GPT-3 with 1000 examples (78%) is surpassed by GPT-4 given only 20 examples (79%). For the third challenge, we again find that GPT-4 outperforms the other two models. It is able to generate problems with accuracy ranging from 76.7% to 100%, depending on the problem type.

We present Blizzard, a Byzantine fault tolerant (BFT) distributed ledger protocol that is aimed at making mobile devices first-class citizens in the consensus process. Blizzard introduces a novel two-tier architecture by having the mobile nodes communicate through online brokers, and includes a decentralized matching scheme to ensure each node connects to a certain number of random brokers. Through mathematical analysis, we derive a guaranteed safety region (i.e., the set of ratios of malicious nodes and malicious brokers for which the safety is assured) for the Blizzard protocol. Liveness is shown as well. We analyze the performance of Blizzard in terms of its throughput, latency, and message complexity. Through experiments based on a software implementation, we show that Blizzard is capable of throughput on the order of several thousand transactions per second per shard and sub-second confirmation latency.

Blockchain technology has the potential to disrupt applications beyond cryptocurrencies. This work applies the concepts of blockchain technology to swarm robotics applications. Swarm robots typically operate in a distributed fashion, wherein the collaboration and coordination between the robots are essential to accomplishing the application goals. However, robot swarms may experience network partitions either due to navigational and communication challenges or in order to perform certain tasks efficiently. We propose a novel protocol, SwarmDAG, that enables the maintenance of a distributed ledger based on the concept of extended virtual synchrony while managing and tolerating network partitions.

Unlike traditional transmission control protocol/Internet protocol–based networks, delay/disruption tolerant networks may experience connectivity disruptions and guarantee no end-to-end connectivity between source and destination. As the popularity of delay/disruption tolerant networks continues to rise, so does the need for a robust and low-latency routing protocol. A one-copy, shortest path delay/disruption tolerant network routing protocol that addresses congestion avoidance and maximizes total network bandwidth utility is crucial to efficient packet delivery in high-load networks and is the motivation behind the development of the congestion avoidance shortest path routing. Congestion avoidance shortest path routing is designed to either route undeliverable packets “closer” to their destinations or hold onto them when advantageous to do so. Congestion avoidance and bottleneck minimization are integrated into its design. Moreover, congestion avoidance shortest path routing negotiates node queue differentials between neighbors similar to backpressure algorithms and maps shortest paths without any direct knowledge of node connectivity outside of its own neighborhood. Simulation results show that congestion avoidance shortest path routing outperforms well-known protocols in terms of packet delivery probability and latency and is still quite efficient in terms of overhead requirements. Finally, we explore the effectiveness of modeling a variant of congestion avoidance shortest path routing based on the resistance to current flow of electrical circuits in an attempt to further reduce network congestion. Preliminary results indicate a noticeable performance increase when this method is used.