Network3 Litepaper
Jan 2024 | Silicon Valley Labs | Version 1.0
This litepaper briefly introduces Network3, a novel Layer1 oriented protocol designed for decentralized, authenticated, anonymous, and reliable data transmission and computation.
Visit our official website to get more information.
1. Introduction
1.1. Problem
In the emerging era of Web3.0, Layer1 leads the way as an open-source, blockchain-based platform, enabling developers to craft and deploy smart contracts and decentralized applications (dApps). However, a gap still exists in the area of decentralized communication for reliable data transmission. This is critical because data exchange impacts the reliability of transmitted transactions, the security of communicating parties, and the overall user experience within the blockchain ecosystem. Therefore, developing a decentralized data transmission protocol that enhances Layer1's scalability and security is of crucial significance.
1.2. Solution
Network3 revolves around key technologies, including an efficient anonymous certificateless signcryption (CLSC) algorithm, a decentralized rating-based data correctness verification mechanism, an IP anti-tracking measure, and a decentralized federated learning framework.
These innovations promise to significantly augment Layer1's Mainnet capabilities and contribute to the ongoing discussion around blockchain openness, security, anonymity, and intelligence. Moreover, the achieved features — anonymity, authenticity, reliability, and decentralized intelligence — set Network3 apart as a groundbreaking solution in the realm of Layer2 protocols.
2. Network3 Contributions
2.1. Efficient Anonymous Authentication
2.1.1. Algorithm Syntax
The foundational cryptographic underpinning of the efficient anonymous authentication protocol in Network3 is the avant-grade certificateless signcryption, which simultaneously provides encryption and signature, thereby ensuring data confidentiality, integrity, authentication. To bolster the protocol’s efficiency, we incorporate the concept of online/offline encryption into the signcryption operations. The key solution to provide immaculate identity anonymity for Layer1 users lies in the construction of a one-time anonymized public key for both communicating parties during each interaction. Attackers are incapable of linking the anonymized public key to any real identity. A smart contract, denoted as SCKGC, supplants a centralized KGC to produce public parameters, generate partial private keys, and distribute them to the corresponding users. Any legitimate user can query public parameters from Layer1.
2.1.2. Concrete Construction and Security Analysis
The entire processes and concrete construction of the proposed anonymous authentication scheme are illustrated in Fig. 1. The proposed authentication scheme can successfully defend against common attacks to ensure security and privacy in blockchain transactions, which has the following potential strengths: 1) anonymity & confidentiality, 2) resistance to KGC compromised attack, 3) replay attack prevention, 4) man-in-the middle attack defense, 5) forgery attack resistance, and 6) protection against big data clustering attack.
2.1.3. Performance Evaluation
We calculated the computational time for all the examined schemes, as illustrated in Figure 2. Notably, Network3 displays remarkable signcryption efficiency in comparison to other schemes, boasting a peak reduction rate of 97%. Its unsigncryption time also stands out favorably. In sum, these findings underscore the practicality and viability of Network3.
2.2. Data Correctness Verification Mechanism
Malicious or selfish data senders might attempt to transmit deceptive or counterfeit data to unsuspecting recipients. Even more severe, toxic data can compromise the system’s normal operation. Therefore, an efficient mechanism for verifying data correctness is crucial.
Network3 has designed a decentralized, privacy-preserving mechanism for verifying data correctness, which leverages the ratings given to transmitted data by stakeholders in the blockchain network. The specific procedures of this mechanism can be divided into seven phases, which are outlined below: 1) Initialization, 2) Rating Submission, 3) Key Shares Dissemination, 4) Key Reconstruction, 5) Average Rating Calculation , 6) Punishment and Incentive, and 7) Shares Update.
2.3. Anonymous Communication Mechanism
2.3.1. Background
As network technology advances, network security, especially the confidentiality of user identities in critical sectors like electronic voting and banking, becomes increasingly important. Despite mature technologies for encrypting communication content, securing the identities of communicating parties is a significant challenge. Existing network protocols like HTTP and TCP/IP are vulnerable to eavesdropping, which can reveal key information such as IP addresses, message length, and packet timing, allowing attackers to deduce user identities and gain valuable insights without needing to access the actual content of the communications.
Current cryptographic methods effectively encrypt content but fail to conceal sender and receiver location information and communication patterns. Attackers can analyze traffic to infer additional information, including address details and communication behaviors, making prominent network nodes and data transmitters potential targets for attacks.
2.3.2. The Tor System and The Optimized Mechanisms
The study focuses on anonymous communication, which aims to hide the communication relationships within business flows without changing existing network protocols, thereby preventing eavesdroppers from discerning or deducing communication links. It begins by examining Tor, a well-known anonymous communication system, and proposes an enhanced mechanism for anonymous communication.
The proposed structure is illustrated in Figure 4, with S symbolizing the sender and R representing the receiver. Sender S1 encrypts the data with the public key of receiver R1and then proceeds to distribute it randomly within the group to which S1 belongs. Once the data reaches S4, it is handed off to the relay network, making its way through the network until it reaches the link’s end. The exit node subsequently transmits the data packet to the multicast group containing receiver R1. All users within this multicast group receive the packet, but only R1, who possesses the private key to decrypt it, can access its contents. All other users can merely discard the encrypted packet, thereby ensuring the privacy of the communication.
2.3.3. Security Analysis
The proposed system is proven to be secure, guaranteeing the overall anonymity of the communication.
1) Low Latency: The implementation of the forwarding network is efficient and straightforward. Data link establishment and data forwarding occur at high speed because they use symmetric keys. The primary factors affecting latency are network bandwidth and congestion. The relay network employs congestion control algorithms and token bucket strategies to ensure network robustness, making the hybrid system achieve very low communication delays.
2) Data Security: In Crowds, any forwarding node can access plain text communication, leaving data security uncertain. In the proposed system, all data in the forwarding network are encrypted by the OP, ensuring data security. It also enables data integrity verification and forward security.
3) Local Eavesdropping: Local eavesdroppers can only observe communications of local users. However, the anonymity of the receiver is maintained because data is disseminated in multicast mode. Local eavesdroppers cannot access any information about the receiver.
4) Multicast Group Eavesdropping: Eavesdroppers within the multicast group may receive the data packet, but it remains encrypted. Only the recipient with the decryption key can access the plain text data. Moreover, the sender’s anonymity is preserved since data packets are forwarded within Crowds, making it difficult for eavesdroppers to ascertain the sender’s information
2.4. Edge Computing-driven Outsourced Computation
Network3 utilized distributed edge computing framework designed to overcome the limitations of terminal equipment like mobile phones and wearable devices, which have limited computational resources and struggle with complex operations. This framework strategically offloads computationally intensive tasks, such as signcryption and key encryption, to edge servers with higher computational capabilities. This offloading significantly reduces the computational load on terminal devices and improves operational performance.
Central to this approach is the use of edge servers, which play a crucial role by handling demanding operations that are transferred from terminal devices. This distribution of tasks ensures efficient operation and enhances the architecture's robustness. The concept of outsourced computation is key here, involving not just task delegation but also ensuring secure computation, data confidentiality, and integrity.
Tasks are allocated to edge servers based on their computational strength, with strict security protocols in place to maintain data confidentiality and integrity. The framework synergistically combines edge computing with outsourced computation, optimizing the use of computational resources, ensuring security, and orchestrating task offloading. This not only addresses the computational limitations of terminal devices but also creates an architecture that is resilient, scalable, and optimized for high performance.
2.5. Decentralized Federated Learning Framework
Network3's solution incorporates several key innovations. Firstly, we integrate layer-2 blockchain technology to decentralize the Federal Learning (FL) processes, mitigating security risks associated with centralized models. Additionally, we introduce an asynchronous local parameter verification mechanism to enhance the robustness of local computations. To further bolster the integrity of the model, a global model verification mechanism is implemented. Our framework also introduces a proof-of-contribution consensus, providing a much more efficient foundation for FL operations. Building upon this consensus, we devise an adaptive incentive mechanism tailored to engage and motivate all participant types in the FL ecosystem. This multifaceted approach aims to not only overcome the limitations of traditional FL frameworks but also establish a more secure, efficient, and participatory paradigm for decentralized machine learning.
2.5.1. Overall Framework and Processes
As illustrated in Figure 5, the proposed framework involves five distinct types of participants. First and foremost, the task publisher, desiring to train an application-specific ML model, releases task descriptions onto blockchain. These descriptions encompass data requirements, convergence demands, rewards, and other pertinent details. Additionally, he/she provides the initial model. Smart contracts are pre-deployed to facilitate task selection, role assignment, and model distribution to trainers. The trainers T use their private dataset to train the local model and subsequently upload the trained model parameters to evaluaters. The local evaluator E are responsible for evaluating the model parameters uploaded by the trainers. The aggregators A play a crucial role in aggregating all legal parameters. The global validators G verify the validity of the global model and generate new blocks. At the end of each round of global model iteration, the adaptive incentive algorithm will allocate corresponding rewards according to the role and work intensity of each node, and update their contribution. At the beginning of the next iteration, Network3 reassigns roles to each node based on its contribution to ensure the liquidity of nodes in the system. All selected nodes engage in model training, parameter evaluation, aggregation, and global validation and block generation tasks, ensuring the timely and positive updating of the global model.
2.5.2. Future Directions
1) Zero Knowledge Machine Learning: Zero-knowledge proofs, a branch of cryptography, have seen significant development in recent years, especially in the field of blockchain. In Layer1, zero-knowledge proofs can be employed for scalability and privacy protection. Layer2 can utilize zero-knowledge proofs to interact with Layer1, ensuring transaction security while enabling off-chain computation to reduce the cost of complex calculations.
2) Model Compression and Pruning: Model compression and pruning have emerged as promising techniques to alleviate communication overheads in distributed federated learning, addressing the challenges posed by transmitting large model updates over communication channels.
3. Conclusion
In conclusion, Network3 presents a groundbreaking Layer1-oriented protocol tailored to address the intricate challenges that have surfaced on Layer1’s Mainnet. By skillfully integrating advanced technologies, including an efficient anonymous certificateless signcryption (CLSC) algorithm, a robust decentralized data correctness verification mechanism, IP anti-tracking measures, and a reliable decentralized federated learning framework, Network3 redefines the landscape of decentralized data transmission and intelligent computing.
The CLSC algorithm stands as a beacon of identity authentication and secure data sharing within an anonymous realm. Network3’s data verification mechanism provides a potent and decentralized solution to the problem of data inaccuracies. Moreover, the proposed anonymous communication mechanism ensures a perfect anonymous pattern for Web3.0 participants. Furthermore, the decentralized FL framework addresses the primary challenges posed by existing FL technologies, providing a promising and practical architecture for intelligent, secure, decentralized computing.
This protocol not only augments the capabilities of Layer1’s Mainnet but also carves a path for the future of DePin ecosystems. It enhances intelligence, bolsters anonymity, and reinforces reliability, setting the stage for an environment conducive to further innovation and progress.
4. Network3 Whitepaper
Last updated
