Investment Thesis — Privasea

Table of contents:

1. Introduction
2. The Importance of Privacy in Computing Infrastructure
3. Enter Privasea
4. Tokenomics
5. Team
6. Partnerships and backers
7. Conclusion

Introduction

With ChatGPT, Open AI started an AI race among the top tech companies in the world. The tools built by these companies solve the same problems in a similar manner. So what gives these companies an edge over the others?

The devils in the data. Any AI tool is as good as the data it’s trained on.

Let’s take the example of ChatGPT and break it down.

ChatGPT is a conversational AI tool designed to revert with the most probable response for any query. GPT = Generative Pre-Trained Transformer. (An Open Source LLM)

A transformer is a computer used for translating data.

• Data is fed into the transformer
• It learns from the data
• It predicts the response
• Finally, with chatbot functionality, we get ChatGPT

To summarize, these LLM models are fed data to recognize and interpret human language or other complex data types, i.e. the quality of the responses is directly proportional to the quality of the data that is fed.

But with lots of data comes a lot of problems. And of the problems is Data Privacy.

The Importance of Privacy in Computing Infrastructure

The race for data has made companies scour the internet to obtain data from all available sources.

You share your data over the internet daily, willingly or unwillingly. Make an online purchase or sign up for an app — with every action on the internet, you share personal information.

While this sea of data has unlocked new possibilities, a data privacy problem has also crept up.

What are the risks of not protecting your personal data?

• Financial frauds
• Identity theft
• Loss of privacy

How do you not hinder the development of AI technology while ensuring data privacy? And why are companies failing to implement privacy solutions:

• Complex Integration: Many enterprises find it challenging to integrate new privacy technologies with existing IT infrastructures.
• Scalability Concerns: Scaling privacy solutions without compromising performance or security remains a technical hurdle.
• Regulatory Compliance: Navigating the complex global landscape of data privacy regulations can be daunting for many organizations.

Fully Homomorphic Encryption seems to be the solution.

Fully Homomorphic Encryption(FHE)

FHE allows you to run computations on encrypted data without decrypting it first.

FHE supports evaluating ciphertext, which outputs an encrypted result that, once decrypted, is similar to running computations on regular, unencrypted data.

What are the primary challenges for FHE?

• Efficiency: Current FHE schemes are slower and require more resources compared to traditional encryption methods
• Not User Friendly: Existing applications require modification, or specialized client-server applications must be developed

Enter PrivaSea

Privasea uses Fully Homomorphic Encryption (FHE), which enables computations to be conducted on encrypted data, producing results that are identical to computations performed on unencrypted data.

What does this mean?

Sensitive data can be processed without putting anyone’s confidentiality at risk.

Privasea has taken it upon itself to adhere to the data regulations that impose rigorous requirements on collecting, processing, and storing personal data. That includes General Data Protection Regulation (GDPR) in the European Union.

By adhering to the strictest regulations and following the best practices for preserving data privacy, Privasea ensures:

• Adherence to the regulatory laws makes people feel safer about their data being processed
• A formidable barrier against data breaches and unauthorized entry by using FHE to encrypt sensitive data during AI processing or inferencing period
• Trust in machine learning systems, encouraging individuals to willingly share their data

How does it work?

Privasea network balances user privacy and distributed computing resources during AI processing. It does so using a comprehensive approach, dividing Fully Homomorphic Encryption from theory to application into four distinct layers:

Application Layer, Optimisation Layer, Arithmetic Layer, and Primitive Layer.

1. With the help of HESea Library, users to encrypt their data or models using a Fully Homomorphic Encryption scheme.
2. Once encrypted, users can securely upload their sensitive data to the Privasea Network and delegate the calculation.
3. Users can perform machine learning or other computations on their data while it remains in an encrypted state.

The Privasea optimized process improves the performance of traditional FHE to a great extent. Here’s a brief outlook:

Roles in Privasea AI Network

Let’s dive into the individual components making the Privasea possible.

Architecture

Source: https://privasea.medium.com/what-is-the-privasea-ai-network-372c00a4837b

• HESea Library: HESea is a collection of highly efficient implementations of popular Fully Homomorphic Encryption schemes like TFHE, CKKS, BGV, BFV, and more. This open-source library equips developers with cryptographic techniques and high-performance optimizations for secure computation. With the HESea Library, developers gain access to a wide range of functions that enable them to perform essential primitive, arithmetic, and logical operations on encrypted data.
• Privasea API: Privasea API is a comprehensive set of protocols and tools built on top of the HESea Library. The API helps developers integrate advanced privacy-preserving features seamlessly into their AI applications.
• Privanetix: Privanetix is an interconnected network of computation nodes enabling secure connections on encrypted data. These nodes leverage FHE algorithms to perform calculations on encrypted data, ensuring that sensitive information remains hidden from the wrong eyes.
• Privasea Smart Contract Kit: The kit includes a series of smart contracts designed to handle various aspects of network management. It incentivizes computation nodes, encouraging their active participation and further strengthening the network’s overall performance
• Accessibility for Non-Cryptographers: One of the significant advantages of Privasea is its user-friendly approach, which allows individuals without a background in cryptography or programming to easily access and utilize the network. This broadens the potential user base and facilitates the adoption of FHE in various applications citation.
• Compliance and Security: The network supports compliance audits and meets various national laws and regulations, including anti-money laundering laws and the EU General Data Protection Regulation (GDPR). This compliance is crucial for users who need to ensure that their data handling practices are legally sound citation.
• Integration with Zama’s TFHE-rs Library: Privasea has integrated Zama’s advanced TFHE-rs library into its network. This collaboration enhances the privacy and security of AI operations within the network, making it a robust solution for data protection(citation).
• Decentralized Computing Network (DePIN): The Privanetix network, driven by Privasea’s smart contract suite, gathers decentralized computing resources to enable secure and efficient processing of encrypted data. This decentralized approach reduces the risks associated with centralized data processing citation.

Let’s take a look at some of the use cases:

1. FHE on Biometrics

2. Proof of Human by FHE

Some additional use cases:

1. KPI Generation: Privasea can use FDFB(Full Domain Functional Bootstrapping) & SIMD(Single Instruction Multiple Data) algorithms to do the statistic analysis in a very efficient way, such as choosing a giant client by measuring KPIs
2. Joint info discovering: Privasea can use PSI(Private Set Intersection) to calculate the intersection parts of 2 datasets without leaking the information to each other
3. Al Modeling: Privasea can use FHE scheme to help banks or insurance groups build a machine-learning model by using the data under encryption to avoid compliance problems

Tokenomics

The PRVA token is the utility token within the network. The demand for the token is directly correlational to the network demand. Let’s understand the token utility to understand the token demand:

Token utility:

1. Transaction Facilitation: PRVA tokens act as a medium of exchange within the network, allowing users to access and pay for privacy AI services offered by Privasea AI
2. Governance & Voting: PRVA holders can participate in the project’s governance by voting on proposals and decisions related to the project using DAO
3. Staking: PRVA can be used for staking, where a certain amount of PRVA can be staked to become a node validator on the Privasea AI Network
4. Exclusive Features: PRVA token may provide users with access to additional features or exclusive benefits within the Privasea AI network. This can include priority access to certain services, discounts, or enhanced functionalities, offering token holders added value and utility within the ecosystem.

Token Distribution (Total Supply: 1B PRVA)

Challenges In Current Landscape & Privasea’s Solution

The current landscape of data privacy and security, especially in the realm of artificial intelligence (AI) and blockchain, presents several challenges. These challenges include computational overhead and efficiency, adoption and practical implementation, privacy and security concerns, regulatory and ethical considerations, and noise management in Fully Homomorphic Encryption (FHE) schemes. Privasea, with its innovative approach and technology, is positioned to address these challenges effectively.

Computational Overhead and Efficiency

Challenge:
FHE schemes, including those that Privasea relies on, are known for their computational intensity and inefficiency compared to traditional encryption methods. This inefficiency stems from complex polynomial calculations and extensive memory interactions due to the large volume of data processed citation.

Privasea’s Solution:
Privasea addresses this challenge by optimizing FHE for better performance. The HESea Library, a collection of highly efficient implementations of popular FHE schemes, equips developers with cryptographic techniques and high-performance optimizations for secure computation1. This optimization improves the practicality and scalability of FHE for real-world applications.

Adoption and Practical Implementation

Challenge:
Despite the theoretical solutions provided by privacy-preserving data mining (PPDM) research, there is little evidence of widespread industry adoption. The barriers include legal requirements, liabilities from inadvertent data disclosure, and the need for sharing information across organizational boundariescitation.

Privasea’s Solution:
Privasea simplifies the adoption of FHE through its comprehensive API and the Privanetix network, which enables secure connections on encrypted data. By making FHE more accessible and user-friendly, Privasea encourages its practical implementation across various industries.

Privacy and Security Concerns

Challenge:
The integration of AI into various sectors raises concerns over data privacy and security. The vast amounts of personal information processed by AI systems necessitate a careful balance between utility and intrusioncitation.

Privasea’s Solution:
By using FHE, Privasea ensures that sensitive data can be processed without compromising confidentiality. This approach builds trust in machine learning systems and encourages individuals to share their data willingly1.

Regulatory and Ethical Considerations

Challenge:
The evolving regulatory landscape, including laws like GDPR, introduces new compliance requirements for organizations. These regulations emphasize the importance of privacy-preserving technologies but also present new challenges17.

Privasea’s Solution:
Privasea adheres to the strictest data protection regulations, such as GDPR, ensuring compliance with legal requirements. This adherence not only makes people feel safer about their data being processed but also positions Privasea as a leader in regulatory compliance1.

Noise Management

Challenge:
Managing noise in FHE schemes is crucial for practical application. Running homomorphic operations increases the noise level in ciphertexts, potentially making them undecryptable1316.

Privasea’s Solution:
While specific strategies for noise management are not detailed in the provided sources, Privasea’s optimization efforts and the development of the HESea Library suggest a focus on improving the efficiency and usability of FHE, which likely includes techniques for effective noise management.

Conclusion

Privasea’s approach to overcoming the challenges in the current landscape involves optimizing FHE for better performance, simplifying its adoption, ensuring regulatory compliance, and enhancing privacy and security. By addressing these challenges head-on, Privasea is well-positioned to lead the AI data revolution, providing secure and efficient data processing solutions for global users citation.

Privasea’s Business Model & Revenue Streams

Subscription Fees:

Privasea charges subscription fees for access to its advanced FHE library and APIs. These subscriptions are likely tiered based on usage levels, providing flexibility for small developers and large enterprises alike.

Transaction Fees:

Fees are charged for computations performed on the Privanetix network. These fees compensate the network nodes for their computational efforts and help maintain the network.

Consulting and Customization Services:

Given the complexity of FHE, Privasea likely offers consulting services to help businesses implement their technology. This could include customization of the HESea Library or integration services for specific business needs.

David Jiao(CEO)

David is the co-founder and CEO of Privasea. A serial entrepreneur with a track record of raising $20M for AI projects and $4M for blockchain initiatives.

Zhuan Cheng

Zhuan is the brains behind everything cryptography at Privasea. He leads the research team to design and develop the cryptographic product architecture and blockchain solutions.

Jeffrey Duan

Jeffrey is a professor and director of Applied Mathematics from the Department of Illinois Institute of Technology. He will be the Chief Science Advisor for Privasea Cryptography algorithm design.

Ting Gao

Associate Professor in HUST. Previous senior ML engineer at Twitter. Chief data scientist of Privasea, who is leading the research team for FHEML.

Lei Chen

A senior researcher and system architect focusing on automotive product architecture and data solutions.

Noel Braganza

With a background in Digital Product Design from the MIT Design Lab, Noel leads the product innovation and design team, focusing on product experience and design strategy.

Sifan Lü

Tech coach for the engineering team. Great experience in building pipeline systems and DevOps.

Partnership & Backers

Privasea’s partnerships with top data infra providers like Chainlink and Filecoin have opened them to a bigger set of audience.

A brief overview of Privasea’s partnerships with live projects:

Privasea X Mind Network:

The expertise in privacy computing infrastructure from Privasea combined with Mind Network’s innovative decentralized data solutions. They are setting new standards for secure, encrypted data exchange, ensuring robust privacy protections in our increasingly digital world.

Privasea X BNB Greenfield:

The partnership combines the robustness of data privacy solution from Privasea and the data storage solution from BNB Greenfield. They are setting a new standard for privacy and control in this data economy.

Further, Privasea’s incubation with Binance Launchpad provides them with a strategic advantage with their BD and marketing efforts.

Roadmap & Milestones

From the start of their journey, Privasea has hit some key milestones:

• Development of the FHE Library: Completion and open-sourcing of the library, making it accessible to developers globally.
• Launch of the Privasea AI Network: Establishing a decentralized network for secure machine learning processes.
• Expansion into New Markets: Targeting sectors like healthcare and finance where data privacy is critical.

What’s ahead for Privasea?

• 2024: Enhance the AI Network capabilities for broader AI model support.
• 2025: Expand global operations, focusing on compliance-driven markets.
• 2026: Introduce next-generation privacy solutions leveraging quantum-resistant encryption technologies.

Conclusion

With multiple advancements in AI technology, the focus will be on data quality. Companies will go above and beyond to gain an edge in the data race. Expanding their sourced data while being compliant and ensuring user privacy is important for these companies. Their only choice is to look for a way to run computations in an encrypted manner without putting the user data at risk.

Privasea offers the world that solution — a network that values privacy. With the focus shifting towards privacy-centric computations, the demand for the PRVA token to access network services can be a positive catalyst for the project.

To summarize, AI tech only advances as far as the data it’s trained on. As long as it’s compliant and ensures privacy, the demand for Privasea Network has the potential to grow exponentially.