With a large number of devices and a wealth of user data sets, the Internet of Things (IoT) has become a great host for federated learning (FL). At the same time, the massive amount of user data in IoT results in desperate demand for privacy preserving. The onion router (Tor) is a promising method to solve the privacy issue in IoT-based FL by user anonymity. However, IoT devices’ resource is too limited to execute the cryptographic operations in Tor. Moreover, network traffics in Tor can be easily controlled by malicious routers with a fake high self-reported bandwidth. In this article, taking advantage of the Tor, we will introduce an anonymous FL framework in IoT called FedTor. To decrease the cryptographic cost in conventional Tor, we propose a lightweight shared key generation scheme for resource-limited IoT devices. Furthermore, we use the difference between the self-reported bandwidth and the bandwidth observed from others to measure the reputation of onion routers. A reputation-based router selection (RBRS) scheme is then brought up to defend traffic control from malicious routers. We conducted extensive simulations to compare FedTor with related works. The results show that the RBRS scheme can decrease the malicious rate of onion routers and the lightweight shared key has a cost advantage over other schemes.

Age of Information (AoI) has become a new metric for data freshness in Unmanned Aerial Vehicle (UAV)-assisted data collection. Whereas, the UAVs are usually energy-limited. When they leave to be charged, the AoI may increase additionally. Current studies of optimizing the AoI usually consider the situation where charging facilities are sufficient, which is not true in many scenarios. For the situation where only one UAV can be charged at one time, we propose a multi-agent cooperative Deep Q-Network algorithm with delayed reward (DR-MADQN) to minimize the AoI. Simulation results show DR-MADQN can effectively decrease the AoI and charge waiting time.

Due to the forecasted fast increasing cellular traffic and the already highly congested licensed spectrum, it is critical to exploit and utilize the unlicensed spectrum resources for the fifth-generation (5G) and beyond networks. A challenging problem is the coexistence of 5G and other networks with fair, reliable, and efficient sharing of the unlicensed spectrum. In this paper, we propose a blockchain-enhanced distributed spectrum sharing scheme for coexisting multiple operators and multiple WiFi APs. We design a novel lightweight and efficient consensus mechanism, named Proof of Strategy (PoG). In this consensus mechanism, the problem of spectrum sharing is used as a consensus puzzle, and the part of the unlicensed spectrum is used as the ‘fee’ of miners. With such a design, the computing overhead of the consensus process is expected to be reduced significantly. We develop a non-cooperative game to analyze the behavior of the miners and obtain a symmetric Bayesian Nash equilibrium under the uniform distribution of mining cost estimation. It can be found mathematically and experimentally that the strategy of the winner tends to maximize the system revenue by sharing the unlicensed spectrum resource. Furthermore, to reduce the impact of heavy interactions on system throughput, the operation of WiFi APs in the proposed scheme can be adaptively switched between ‘contention mode’ and ‘blockchain mode’ according to the network traffic load. The dynamic behavior is constructed as an evolutionary game, and the existence and uniqueness of equilibrium points are proved by theoretical analysis. Simulations demonstrated the fairness and effectiveness of the proposed blockchain-based scheme and the mode switching method for distributed spectrum sharing by heterogeneous wireless networks.

Machine learning is an essential technology providing ubiquitous intelligence in Internet of Things (IoT). However, the model training in machine learning demands tremendous computing resource, bringing heavy burden to the IoT devices. Meanwhile, in the Proof-of-Work (PoW)-based blockchains, miners have to devote large amount of computing resource to compete for generating valid blocks, which is frequently disputed for tremendous computing resource waste. To address this dilemma, we propose an Evolved-PoW (E-PoW) consensus that can integrate the matrix computations in machine learning into the process of blockchain mining. The integrated architecture, the elaborated schemes of transferring matrix computations from machine learning to blockchain mining, and the reward adjustment scheme to affect the activity of the miners are, respectively, designed for E-PoW in detail. E-PoW can keep the advantages of PoW in blockchain and simultaneously salvage the computing power of the miners for the model training in machine learning. We conduct experiments to verify the availability and effect of E-PoW. The experimental results show that E-PoW can salvage by up to 80% computing power from pure blockchain mining for parallel model training in machine learning.

Distributed machine learning (DML) can realize massive dataset training when no single node can work out the accurate results within an acceptable time. However, this will inevitably expose more potential targets to attackers compared with the non-distributed environment. In this paper, we classify DML into basic-DML and semi-DML. In basic-DML, the center server dispatches learning tasks to distributed machines and aggregates their learning results. While in semi-DML, the center server further devotes resources into dataset learning in addition to its duty in basic-DML. We firstly put forward a novel data poison detection scheme for basic-DML, which utilizes a cross-learning mechanism to find out the poisoned data. We prove that the proposed cross-learning mechanism would generate training loops, based on which a mathematical model is established to find the optimal number of training loops. Then, for semi-DML, we present an improved data poison detection scheme to provide better learning protection with the aid of the central resource. To efficiently utilize the system resources, an optimal resource allocation approach is developed. Simulation results show that the proposed scheme can significantly improve the accuracy of the final model by up to 20% for support vector machine and 60% for logistic regression in the basic-DML scenario. Moreover, in the semi-DML scenario, the improved data poison detection scheme with optimal resource allocation can decrease the wasted resources for 20-100%.

The Proof-of-Work (PoW) consensus is widely used to maintain the security and consistency of the blockchain systems. In PoW, the miners have to consume tremendous computing resources to compete for generating a valid block and win the reward. However, in recent years, the revenue of the miners face a large oscillation, due to economic policy changes, virtual currency price fluctuations, etc. Some miners have to stop mining and exit the market, leaving large number of computing resources vacant. Whereas, vast computing resources are required outside the blockchain ecosystem, such as cloud/edge computing and artificial intelligence. In this paper, the traditional miners (T-miner) which concentrate on block mining are promoted to the evolutionary miners (E-miner) with additional ability to provide computing services. We propose a scheme for the E-miners to connect the blockchain ecosystem with external computation requirements, releasing the redundant computing power of the E-miners to provide computing services for extra profits. A two-stage Stackelberg game is presented to model the interaction between the computing service pricing and each E-miner’s computing power allocation. The details of the game are separately elaborated for both flexible and specified external computing requirements, in pure environment with only E-miners and hybrid environment with both E-miners and T-miners, respectively. We utilize two distributed iteration algorithms to find the equilibrium point of the Stackelberg game, where both sides can achieve optimal utility. Extensive simulation results demonstrate the effect of our proposed schemes, and show that the profits of the E-miners are considerably increased compared with the T-miners that devote all computing power to block mining.