Arcium specializes in encrypted computations, allowing different people or businesses to work together on tasks while keeping their data private safely.

At the core of this is Multi-Party Computation (MPC), an advanced encryption technique that enables multiple parties to collaborate on computing tasks while maintaining the privacy of their data. With MPC, users can securely share and compute data without exposing sensitive information. It’s like a group working on a puzzle where each person contributes a piece, but the final result is the only thing visible. For a deeper dive into how MPC works, check out our MPC ELI5 piece.

Arcium uses two different MPC models, or security assumptions, for these calculations. One is designed to prioritize speed, while the other emphasizes security. In combination, they allow for nearly any type of use case to be supported across both Web2 and Web3.

The two security approaches are:

1. Cerberus: Strong security and privacy guarantees, perfect for sensitive data such as financial transactions. It operates under a “dishonest majority” model, meaning it works even if the majority of the network is malicious.
2. Manticore: A faster system designed for compute-heavy applications such as AI and machine learning modeling. It operates in an “honest but curious” model which assumes most parties are honest, meaning faster computations.

In this article, we'll explore the differences between these two and which use cases they're best for.

Understanding both security approaches is crucial as it helps strike the perfect balance between security and performance, whether you're building secure DeFi transactions, training AI models, or anything else that’s possible with MPC.

MPC: A Quick Primer

Although we covered MPC in our MPC ELI5 piece, to provide a high-level recap, it’s important to note that MPC is a privacy-preserving technology that allows multiple parties to collaborate on computations without exposing their private data. 

It’s a key enabler of Privacy 2.0, allowing secure, private collaboration between multiple parties and applications without exposing sensitive data. This contrasts with Privacy 1.0, where privacy was limited to isolated, single-user use cases.

MPC is also very fast and highly flexible, capable of serving many Web2 or Web3 use cases.

All in all, this can be applied to a wide variety of Web2 or Web3 use cases, but there are specific MPC, which is where Cerberus and Manticore come in.

‍Cerberus

Cerberus is Arcium's main approach to protecting data while it’s being computed upon. It operates on a "dishonest majority" model, which means it can still work securely as long as a single network participant is honest, known as an “N-1” security assumption.

Here’s how it works:

• Security First: Cerberus is designed to stop bad actors. If any party tries to cheat or break the rules, the system stops the computation immediately. Unlike other MPC protocols that require most parties to be honest, Cerberus only needs one honest participant to keep things secure.
• No Trusted Dealer: Cerberus doesn’t rely on a single party to share cryptographic secrets. Other systems often have a "trusted dealer" that can be a target for attacks.
• Verification System: Cerberus uses a security feature called Message Authentication Code (MAC) to ensure that every participant’s data is correct. If anything is wrong, Cerberus stops the process, and the faulty party is replaced.
• Faster Computation: Cerberus accelerates operations by performing additions and multiplications on encrypted data without the need for decryption. This ensures quick processing while preserving privacy, similar to Somewhat-Homomorphic Encryption (SHE), which enables specific calculations on encrypted data without exposing it.

Think of Cerberus like a quality control system in a factory. There are multiple inspectors (nodes) checking the work at the same time. If any honest inspector notices a problem, the whole process stops immediately, ensuring everything is secure.

Cerberus combines strong security with fast processing, and is best for permissionless settings where network participants are anonymous and may be malicious. Watch Co-founder Yannik Schrade talk about Cerberus here:

Manticore

Manticore is Arcium’s fastest MPC protocol, operating under the honest-but-curious model, which is a variant of the honest majority approach. In this model, participants follow the protocol but might attempt to gain additional information beyond what is necessary. 

This approach is ideal for permissioned settings, where trusted operators are involved. It’s therefore very fast, as it assumes that all participants are honest.

It shares a few key differences with Cerberus, notably:

• Speed Over Security: Manticore is designed for speed, so it can handle large-scale computations quickly. However, it comes with a trade-off in security.  Unlike Cerberus, which ensures high security, Manticore operates under the "honest but curious" assumption. This means it assumes the participants will act honestly, but they may still try to learn more about the data than they should, hence the "curious" part. 
• Trusted Environment: Manticore uses a "trusted dealer" model, where one party is responsible for generating cryptographic secrets. While this speeds up the process, it introduces a potential vulnerability because the dealer must be trusted, though this risk is minimized as Manticore is typically used in controlled environments.

Think of Manticore as a high-speed immigration lane at the airport, for those who have been pre-cleared by security. While the regular immigration lane may have more security (like Cerberus), the pre-clearance lane still has a degree of security, but is faster as it assumes that passengers are trustworthy.

Bringing Things Together

Both Cerberus and Manticore are advanced encryption protocols that enable multiple parties to collaborate securely on computations while ensuring the privacy of their data. By allowing participants to collaborate without exposing sensitive information, these protocols unlock a wide range of possibilities across Web2 and Web3 ecosystems.

The full extent of applications where Cerberus and Manticore can be used is vast, as both protocols are flexible and designed to meet the security and performance needs of various use cases. Whether it’s DeFi, healthcare, AI model training, or beyond, both Cerberus and Manticore are optimized for different scenarios where privacy is crucial.

Here are a few examples of how they can make an impact:

• DeFi: Traders can execute transactions in dark pools, where their strategies and positions remain confidential, avoiding exposure to competitors.
• Healthcare Data Collaboration: Institutions can share encrypted medical data for research or diagnosis, protecting patient confidentiality while collaborating on innovative treatments.‍
• AI Model Training: Companies can train AI models on encrypted customer or financial data, ensuring privacy without compromising the value of the data.

While Cerberus provides a higher level of security through its multiple rounds of verification, this comes at a slight expense in terms of speed. On the other hand, Manticore focuses on speed by assuming that participants act honestly, which makes it ideal for more complex computations that require fast processing.

To dive deeper into Arcium’s advanced encryption protocols and understand how they enable secure and private computations, we recommend:

• Reading our recently released Purplepaper, which provides a comprehensive breakdown of Arcium’s system and its cutting-edge approach to privacy-preserving computations.
  
• Exploring the Arcium Docs, a detailed guide for developers looking to integrate Arcium’s technology into their applications, ensuring seamless adoption and implementation.

‍