Digestive neural networks: A novel defense strategy against inference attacks in federated learning
نویسندگان
چکیده
Federated Learning (FL) is an efficient and secure machine learning technique designed for decentralized computing systems such as fog edge computing. Its process employs frequent communications the participating local devices send updates, either gradients or parameters of their models, to a central server that aggregates them redistributes new weights devices. In FL, private data does not leave individual devices, thus, rendered robust solution in terms privacy preservation. However, recently introduced membership inference attacks pose critical threat impeccability FL mechanisms. By eavesdropping only on updates transferring center server, these can recover device. A prevalent against differential scheme augments sufficient amount noise each update hinder recovering process. it suffers from significant sacrifice classification accuracy FL. To effectively alleviate problem, this paper proposes Digestive Neural Network (DNN), independent neural network attached The owned by device will pass through DNN then train modifies input data, which results distorting way maximize while minimized. Our simulation result shows proposed performance both gradient sharing- weight sharing-based For sharing, achieved higher 16.17% 9% lower attack than existing schemes. sharing scheme, at most 46.68% success rate with 3% accuracy.
منابع مشابه
Reinforcement Learning in Neural Networks: A Survey
In recent years, researches on reinforcement learning (RL) have focused on bridging the gap between adaptive optimal control and bio-inspired learning techniques. Neural network reinforcement learning (NNRL) is among the most popular algorithms in the RL framework. The advantage of using neural networks enables the RL to search for optimal policies more efficiently in several real-life applicat...
متن کاملReinforcement Learning in Neural Networks: A Survey
In recent years, researches on reinforcement learning (RL) have focused on bridging the gap between adaptive optimal control and bio-inspired learning techniques. Neural network reinforcement learning (NNRL) is among the most popular algorithms in the RL framework. The advantage of using neural networks enables the RL to search for optimal policies more efficiently in several real-life applicat...
متن کاملPacket Leashes: A Defense against Wormhole Attacks in Wireless Networks
As mobile ad hoc network applications are deployed, security emerges as a central requirement. In this paper, we introduce the wormhole attack, a severe attack in ad hoc networks that is particularly challenging to defend against. The wormhole attack is possible even if the attacker has not compromised any hosts, and even if all communication provides authenticity and confidentiality. In the wo...
متن کاملDefense against Routing Disruption Attacks in Mobile Ad Hoc Networks
We propose a secure routing architecture for Mobile Ad hoc NETworks (MANETs) called ThroughpUt-Feedback (TUF) routing, which is resilient against a wide range of routing disruption Denial-of-Service (DoS) attacks. Unlike many existing solutions, TUF does not focus on a particular type of attack, but instead takes an approach that is fundamentally more general. TUF is a cross-layer technique tha...
متن کاملSequential Defense Against Random and Intentional Attacks in Complex Networks
Network robustness against attacks is one of the most fundamental researches in network science as it is closely associated with the reliability and functionality of various networking paradigms. However, despite the study on intrinsic topological vulnerabilities to node removals, little is known on the network robustness when network defense mechanisms are implemented, especially for networked...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Computers & Security
سال: 2021
ISSN: ['0167-4048', '1872-6208']
DOI: https://doi.org/10.1016/j.cose.2021.102378