write e

FL in NLP

1. Federated Learning of Unsegmented Chinese Text Recognition Model:

In its current stage, public federated learning frameworks still suffer from series of errors and lacking features for deep model training. For one thing, FL frameworks should fulfill the implementation with GPU support, so as to harness the computation power of fast-increasing computation power of GPUs. Secondly, in current configurations, the workload of a federated learning system mainly lies on client servers, where iterative gradient computation and model update are conducted. Thirdly, most current researches in the federated learning literature focus on machine learning methods, such as SVM, logistic regression and linear regression, etc. It would be more interesting to see how federated learning enables training of new models that was hindered by the lack of data collection.

https://ieeexplore.ieee.org/abstract/document/8995347

FL in autonomous vehicles

1. Distributed Federated Learning for Ultra-Reliable Low-Latency Vehicular Communications:

In this regard, a future extension of this work is to study FL in the presence of non-IID training data.

https://arxiv.org/pdf/1807.08127.pdf

2. Federated Learning for UAVs-Enabled Wireless Networks: Use Cases, Challenges, and Open Problems:

https://ieeexplore.ieee.org/document/9039589?denied=

3. Improving TCP Performance Over WiFi for Internet of Vehicles: A Federated Learning Approach:

Detailed analysis of the convergence conditions for our IoV system dynamics with proposed algorithm requires further investigations and is ongoing.

https://d1wqtxts1xzle7.cloudfront.net/62911641/IEEE_LetterShivaJinho-1820200411-11554-1irjm4z.pdf?1586602342=&response-content-disposition=inline%3B+filename%3DImproving_TCP_Performance_over_WiFi_for.pdf&Expires=1595175416&Signature=dVji-Kd6i7P7vGXiSA8B0IQ2I5kLDmF1fOjemt1-aO4Ka57-CgMwaktT92H8V~acmzw7tGqVIZOf1rDcKAw-nsDDnPLMR8KkcfkkndvxGtfkFsMD7RMctomscizlXlMATzJY3gUTc5QOoIm7E3UqZFC5D2dgvhalg4ZhPCw76f-5GT0L8RSF30JcY4Ivee1sdPqDZe28n~PPncjq4VZREnIk62vVVq1HwFMyw3wqN30fcGED2Z05xgpUtbelG2HozBq6mAdkHjUS-p1boaILwsDfSCJaOxx0N5vqqSx~1iCsu-dGMOX0~1HWbATPb3vnu3KSmBNqa0~CA6lBqY-M-Q__&Key-Pair-Id=APKAJLOHF5GGSLRBV4ZA

4. Federated Learning with Blockchain for Autonomous Vehicles: Analysis and Design Challenges:

Finally, we conclude with a summary of the simulation and numerical findings. We have observed that there are optimal pairs (?;out) in the sense that the system delay E(n;m) [T] is minimal with that value of (;out). Clearly, on one hand, one should use a sufficiently large number of retransmission to combat transmission failures, but on the other hand R should not be so large that the repeated transmissions and the resulting service time jointly increase(n;m)[T]. Also important is the choice of an optimal block/update size, which will limit the increase in the overall delay but; a detailed study of this effect will be the topic of our future research.

5. Federated Learning for Data Privacy Preservation in Vehicular Cyber-Physical Systems:

Numerical results on a real-world dataset show that the proposed scheme can mitigate data leakage in VCPS effectively. Federated learning is an emerging paradigm for privacy-preserving training over distributed data. However, clients still have a chance to learn the private data of others by collecting or eaves-dropping the model parameters. How to enhance data privacy in federated learning remains an open issue. Moreover, developing theoretical or analytical frameworks for deriving the secrecy rate is helpful to assess how data privacy is preserved in federating learning for VCPS. Also, the application of differential privacy can be further explored.

FL for Power Management

1. Incentive Design for Efficient Federated Learning in Mobile Networks: A Contract Theory Approach:

For further work, we will consider using blockchains to ensure the reliability of local model updates when formulating the incentive mechanism for reliable federated learning in mobile networks [17], [18], [19], [20]

FL Privacy

1. Privacy-Preserving Asynchronous Federated Learning Mechanism for Edge Network Computing:

As a relatively new research content, asynchronous learning still has much room for discussion. As mentioned above, asynchronization is caused by many factors, and different causes require different solutions. In future work, we will discuss different attenuation functions and look for a better attenuation function to fit the attenuation requirements.

2. Privacy-aware service placement for mobile edge computing via federated learning:

For future work, we will consider using PSP for several edge clouds.

3. PDGAN: A Novel Poisoning Defense Method in Federated Learning Using Generative Adversarial Network:

In future work, we plan to explore the poisoning defense for federated learning with device, class, or user-level differential privacy

FL in IoT

1. CoLearn: enabling federated learning in MUD-compliant IoT edge networks:

In the current CoLearn deployment, we assumed that edge devices do not fail in the training phases and during their activity of traffic eavesdropping. This assumption necessitates further considerations and improvements. Additionally, we did not focus on IoT device identification and authentication, which is vital for both MUD-compliant networks and FL architecture, and can be solved by using IoT device manufacturer provisioned X.509 certificate.

2. Distributed Sensing Using Smart End-User Devices: Pathway to Federated Learning for Autonomous IoT:

In future work, we plan to develop distributed learning algorithms tailored for AIoT (Autonomous Internet of Things) applications.

3. Multi-Agent DDPG-based Deep Learning for Smart Ocean Federated Learning IoT Networks:

As future research directions, we will implement and deploy this proposed MADDPG-based multi-agent DRL algorithm in real-world scenarios under the consideration of specific FL dataset and applications. Furthermore, data-intensive measurement-based performance evaluation can be done with various measurement metrics.

4. Real-Time Data Processing Architecture for Multi-Robots Based on Differential Federated Learning

In our future work, in addition to the robotic recognition work in the real scenario, we will apply our architecture into more real-time IoT applications.

5. Federated Learning for Vehicular Internet of Things: Recent Advances and Open Issues

We then discussed the future research directions on the integration of FL with vehicular IoT taking into account both the application of FL for vehicular IoT, and the enhancement of vehicular IoT technologies for supporting FL. We believe this work could expedite the research process for both FL and vehicular IoT.

6. Distributed Federated Learning for Ultra-Reliable Low-Latency Vehicular Communications:

In this regard, a future extension of this work is to study FL in the presence of non-IID training data.

FL Privacy-preserving

1. HybridAlpha: An Efficient Approach for Privacy-Preserving Federated Learning:

Most of the existing federated learning frameworks only work on the scenario of horizontally partitioned data. To tackle the issues and challenges in the case of vertically partitioned data, several methods are proposed in [9, 14, 22], which focus on entity resolution and simple machine learning models, like logistic regression. Such vertically partitioned data cases beyond the research scope in this paper and will be deferred to future work.

2. PMF: A Privacy-preserving Human Mobility Prediction Framework via Federated Learning:

The future work of our work can be divided into two directions: 1) designing advanced mobility model for better personal modelling. In this paper, we only consider the basic mobility model for the simplicity of the whole system. 2) expanding our framework to more types of machine learning models and different scenarios.

3. PrivFL: Practical Privacy-preserving Federated Regressions on High-dimensional Data over Mobile Networks:

As a future work, it would be interesting to investigate the case of users asynchronously participating in the training phase. Acknowledgement. This work is supported by the NSERC Discovery grant. The authors would like to thank the anonymous reviewers of CCSW2019 for their insightful comments and suggestions to improve the quality of the paper.

4. Highly efficient federated learning with strong privacy preservation in cloud computing:

Our approach also opens up several directions for future research. For example, instead of only using SMC techniques to prevent inference during the training process, combining SMC with DP can further prevent inference over the outputs. Besides, it is also interesting to explore other SMC techniques such as Pallier instead of ElGamal, which may further enhance the training efficiency.

5. A training-integrity privacy-preserving federated learning scheme with trusted execution environment:

Therefore, we can say that our scheme can prevent causative attacks in federated learning such that the integrity of the well-trained global model is protected. Alternatively, the data verification problem is still an interesting and important topic, and we can leave this problem for future researches in the federated learning.

6. A unified data security framework for federated prognostics and health management in smart manufacturing:

The future work will include designing a solution for automatically determining the working regime and sharing it among the network agents without violating privacy.

7. EaSTFLy: Efficient and secure ternary federated learning:

In the future, on the one hand, we would like to improve the efficiency of privacy-preserving federated learning, and on the other hand, we will focus on resisting the more powerful adversary.

8. Privacy-preserving federated k-means for proactive caching in next generation cellular networks:

In our future work, we will implement PFK-Means with the programming language that has higher performance than Python, e.g., C/C++. With regard to the high computational cost caused by data reconstruction in PFK-Means, we will further 18 investigate how to reconstruct dropout user’s data in a more efficient way. Furthermore, if a more powerful computation platform is deployed, PFK-Means can obtain more excellent performance.

9. Privacy-Preserving Federated Learning in Fog Computing:

Our future work can be devoted to improving the efficiency of our scheme and reducing the computational cost, which makes it more practical and maintains security.

FL Privacy concerns

1. FFD: A Federated Learning Based Method for Credit Card Fraud Detection:

In future works, we will take more reliable measurements into account to protect the privacy of data. And the Non-IID dataset can be evaluated in this credit card fraud detection system and ensure the credit card FDS to communicate and aggregate the model updates in a secure, efficient and scalable way.

2. Preserving Data Privacy via Federated Learning: Challenges and Solutions:

Some applications related to IoT ecosystem, genome-studies, smart city, and finance application via federated learning can be the candidate areas for future works.

FL in Health Care

1. FedHealth: A Federated Transfer Learning Framework for Wearable Healthcare:

FedHealth opens a new door for future research in wearable healthcare. In the future, we plan to extend FedHealth to the detection of Parkinson’s disease where it can be deployed in hospitals.