3.2 Technical architecture system

3.2.1 Hardware layer: encrypted security environment and data collection terminal

(1) Core hardware configuration

  • AI computing center:

The Dimensity 8200 octa-core processor (4nm process, 1 3.1GHz Cortex-A78 large core, 3 3.0GHz Cortex-A78 large cores, and 4 2.0GHz Cortex-A55 small cores) optimizes neural network computing performance through instruction-level dynamic scheduling. Its integrated fifth-generation APU 580 supports INT8/INT16 mixed precision operations, and its AI computing power is about 8TOPS. With UFS 3.1 storage and LPDDR5X memory, the system can process multiple AI tasks in parallel (such as real-time transaction signature verification and multi-chain wallet status synchronization), and ensure the independence and stability of key tasks through memory domain isolation technology.

  • Multi-camera spatial modeling system

The 100-megapixel rear main camera (OIS optical image stabilization) and the 32-megapixel front camera form a multi-line stereo vision system. Through the deep learning parallax estimation algorithm, sub-centimeter depth perception is achieved within the range of 0.5-4 meters. Innovative fusion of IMU inertial measurement unit data (sampling rate 1000Hz) dynamically compensates for the impact of hand-held jitter on spatial modeling. In DeFi application scenarios, the system can provide stable spatial anchoring capabilities for AR asset displays in physical stores.

  • Quantum security engine

The under-screen optical fingerprint module adopts multi-spectral biometric technology:

① Liveness detection: 780nm short-wave infrared light penetrates the epidermis to collect the capillary topology of the dermis (rejects silicone fingerprint film attacks).

② Encrypted link: The fingerprint feature generates a 320-bit quantum-resistant hash value through the national secret SM3 algorithm and stores it in the physical detection-proof memory of the TEE security zone (PUF technology realizes the binding of keys and chips).

③ Performance indicators: The unlocking speed in dry and wet finger scenarios is 0.28 seconds, FAR (false acceptance rate) <0.002%, and supports 100,000 consecutive authentications without attenuation. Trusted Execution Environment (GlobalPlatform Certified TEE)

In the three-layer protection system built by ARM TrustZone, the Cosmic Cipher series of products supports closed-loop protection from key generation to destruction at the hardware level:

① Secure generation: Integrates a true random number generator that complies with the NIST SP 800-90C standard, with an entropy value of ≥0.97 per bit.

② Hardware binding: The private key is stored in the OTP storage area of ​​the eSE security chip and is bound to the physical fingerprint of the processor (to prevent transplant attacks).

③ Policy execution: Each key is associated with a double factor of application hash value + fingerprint feature, for example, the BTC wallet key only responds to Bitcoin Core client calls that have passed the signature verification.

  • Transaction acceleration engine

The Cosmic Cipher series of products supports optimization of blockchain operations through heterogeneous computing architecture (can be optimized through hot update system): a.CPU acceleration: The Dimensity 8200 processor has a main frequency of 3.35GHz, supports a dedicated encryption pipeline, and the secp256r1 curve signature speed can reach 412 times/second (19% increase over the previous generation). b.Storage optimization: Supports UFS 3.1 storage, enables Turbo Write technology, and the mnemonic encryption backup rate is increased to 1.2GB/s (full disk erase takes <2.8 seconds). c.Network collaboration: 5G baseband integrates IPSec hardware acceleration, and block broadcast delay is reduced to 18ms (73% increase compared to LTE environment).

(2)Biometric Trust Chain

Single biological factor deep binding protocol:

Registration stage:

  1. Multispectral fingerprint scanning generates a 384-bit root key (following the BIP-44 standard)

  2. The characteristic data is encrypted by the SM2 algorithm and stored in the tamper-proof storage area of ​​the security chip

Transaction phase:

  1. Fingerprint press activates the key container (response time <300ms)

  2. Dynamically extract the changes in fingertip micro-textures to generate random salt values ​​(anti-side channel attacks)

  3. Hash summary participates in the construction of Schnorr multi-signature

(3)Liveness detection matrix

  1. Optical topological analysis:

    Capturing the three-dimensional structure of fingerprint ridges through a 3D macro lens (rejecting 2D printed images)

  2. Blood flow imaging: infrared spectroscopy to detect the dynamic distribution of hemoglobin in fingertips (rejecting static biological models)

  3. Behavioral feature learning: record pressure and contact surface shape to establish a user-specific bio-behavioral profile

3.2.2 Visual layer: 3D creation kit - 3D light field display native content engine

On the visual level, Cosmic Ciphe uses advanced light field display technology to achieve the following on a 6.95-inch AMOLED screen:

  • Dynamic parallax compensation:

The innovation of light field display technology lies in its adaptive optical properties. Each square centimeter contains 1,200 independently controlled cylindrical lenses. When the built-in gyroscope detects the movement of the user's head, the FPGA controller will recalculate the lens curvature within 3ms so that the screen pixel light is accurately directed to the pupils of both eyes. With a 120Hz refresh rate, a dizziness-free stereoscopic visual experience is achieved. In the game scene, when the player observes sideways, the system automatically adjusts the depth of field parameters to keep the spatial position of the virtual object stable.

  • Multi-format adaptation:

For 3D content from different sources, the display system has built-in dedicated decoding chips: for left-right format videos of movies and TV shows, the frame rate is increased to native 120fps through AI interpolation technology; for professionally produced 2D+Z depth map formats, the depth information is used to reconstruct the stereoscopic parallax field; the most innovative is the automatic recognition function of multi-grid formats - when it is detected that the picture is divided into 9 grids, the system will extract the parallax parameters of each block and dynamically synthesize a 360-degree panoramic effect. These conversion processes are all completed in the display driver chip to ensure that the picture delay is lower than the threshold of human visual perception.

  • Eye comfort mode:

On the basis of traditional blue light filtering, the system tracks the user's pupil contraction status in real time through the front camera. When it detects that the user has been staring at the screen for more than 20 minutes, it will automatically start a composite protection strategy: first adjust the refractive index of the liquid crystal lens to shift the focus from the screen plane to a virtual distance of 1 meter; at the same time, dynamically filter the 415-455nm band blue light and adjust the color temperature to a warm 5500K. Clinical tests have shown that this technology can reduce the visual fatigue index by 62%.

At the same time, by taking advantage of this hardware feature, a full-link solution of "creation-optimization-presentation" was built to reshape the mobile terminal 3D content display scene through three major scenarios:

(1)2D to 3D cloud platform: 3D light field display native content generation

  • Scenario-based solutions

  1. 3D product: Merchants upload 2D images (such as shoes/jewelry), and the system automatically generates a 3D light field model with depth information.

  2. Interactive display: users can view product details 360° by tilting their phone left or right (using gyroscope data to dynamically adjust depth of field)

  3. Data enhancement:

- Optimize the model normal map through the microlens array to eliminate moiré

- Automatically adjust the model face density according to screen parameters (matching 1240*2772 pixel arrangement)

  • Technological breakthrough

3D light field display technology adaptation algorithm:

- Automatically optimizes model polygon density to 1/4 of the screen lens spacing (anti-aliasing)

- Dynamically adjust parallax parameters (0.5-2.5 degrees), suitable for 30cm-80cm viewing distance

(2)Real-time rendering engine:3D light field display optimization

  • Lens compensation rendering:

According to the optical parameters of light field display technology, the rendering perspective is corrected in real time.

  • Multi-focal plane synthesis:

Render the scene in layers (foreground/midground/background) using dynamic zoom technology:

- Foreground objects (distance <30cm): 1080P@90Hz high refresh rate

- Background environment (distance>1m): 720P@45Hz low power mode

  • Performance optimization

Optimization technology
effect improvement
hardware linkage

Parallax-aware LOD

Rendering load reduced by 40%

Calling gyroscope data to dynamically simplify the model

Lens Aligned Anti-Aliasing

92% reduction in moiré

GPU Microlens curvature data is fed back to the GPU in real time

Ambient light adaptation

Dark details increased by 3 stops

Light sensor linkage to adjust HDR parameters

(3)Multimodal synthesis workshop: 3D light field dynamic narrative

Case scenarios, such as film and television three-dimensional short videos:

A. Dual channel generation:

- Runway channel generates basic animation (character action/scene transformation)

- Pika channel adds 3D light field effects (particle spot/deep field fog)

B. Mobile optimization:

- Automatically map video depth range to the screen's best perception interval (2D+Z to multi-view format)

- Dynamically adjust the stereo intensity according to the ambient light intensity (intensify 20% parallax under strong light)

(4)Panoramic view of scene solution

3.2.3 Blockchain layer: Contribution value rights confirmation system

The blockchain layer aims to build a decentralized value rights confirmation system that transforms users' participation in AI services (data contributions, API calls, etc.) into quantifiable on-chain assets. Ensure privacy and security through zero-knowledge proof (ZKP), anchor user identity through digital human NFT, and ultimately form a closed-loop of value of "behavior-points-tokens". The entire system contains five core modules.

(1)Data contribution collection case process

type
Source of data
Value weight (assumption)
Sample Scenario

Digital Human Interaction Contribution

API call log

0.8

Digitally generated pictures/text

Data contribution

Sensor/user input data

1.2

Upload photos for model training

Network maintenance contribution

The online duration of the device

0.5

Provide data transfer to other users

(2)Local data encapsulation

The device side is completed through TEE (Trustable Execution Environment)

  1. Data desensitization: remove personal identity information (PII) and retain characteristic dimensions

  2. Feature extraction: convert the original data into a standardized format (such as converting the picture into a 128-dimensional feature vector)

  3. Hashing evidence storage: generate data fingerprint (SHA3-256) and temporarily store it in the local security area

(3)Zero Knowledge Proof (ZKP) workflow

Use the zk-STARKs protocol to generate contribution credentials:

Circuit compilation: convert contribution verification logic into arithmetic circuit

Constraint: Contribution value ≤ physical upper limit (such as the maximum number of API calls per day for mobile phones)

Proof generation: Generate proof documents within TEE (about 50KB)

On-chain verification: Verification nodes verify validity through smart contract verification

(4)Points Casting and Distribution

  • Integral calculation formula:

Integral = Σ (contribution type weight × contribution amount × device coefficient)

· Equipment coefficient: Based on hardware performance grading (assuming flagship model = 1.3, mid-range model = 0.7)

· Attenuation factor: Daily efficiency decreases by 3% when continual contribution (prevention of brush volume)

  • On-chain casting process

  1. User submits zk-STARKs certificate to smart contract

  2. Contract verification proves validity (call off-chain oracle assistance)

  3. Cast corresponding points to the SBT NFT bound by the user

  4. Update NFT metadata (total points/historical contribution record)

(5)Token rewards are issued

  1. The user initiates a request for points redemption of MMI tokens (signature transaction)

  2. Contract destroys corresponding points (update SBT status)

  3. Calculate the amount of MMI to be issued based on the current exchange rate

  4. Transfer tokens from the fund pool to the user's wallet

  5. Trigger token destruction (5% MMI is destroyed by a single transaction)

(6)Governance and Extension Interfaces

  • DAO governance module

① Adjustment of points weight

② Change of redemption rate parameters

③ Access to cooperative API service providers

· Voting mechanism

a. 1 veMMI = 1 vote

b. Two-factor certification is required for major decisions (veMMI+Mind Mosaic Identity SBT hold)

  • Partner collaboration interface

Reserve standardized API for external computing power suppliers to access:

  1. Release of computing power tasks: The mine can define computing tasks that require mobile phone assistance.

  2. Contribution verification protocol: Use improved version of PoRA (Proof of Resource Allocation)

  3. Revenue distribution contract: Automatically split the mine income to participating equipment

(7)Privacy Computing Layer

  • Secure multi-party computing (MPC)

- Coordinate the global contribution statistics between devices (such as the average daily API calls of the entire network)

- Optimize communication efficiency with Beaver triple

  • Federated Learning Framework

- Local model training only uploads gradient parameters (differential privacy protection, ε=0.5)

- Gradient aggregation smart contracts verify data validity

(8)Storage and verification layer

  • Decentralized storage

- Raw data: IPFS shard storage (CID hash is recorded on the chain)

- Verification data: Arweave permanent archive

  • Oracle network

Integrate Chainlink two-factor verification mechanism:

- Off-chain computing: Verify complex ZKP proof

- Data feeding: Obtain the real-time exchange rate of MMI/USDT

(9)Witch Attack Defense

  • Equipment fingerprint system

Fusion triple hardware features to generate a unique ID:

Fingerprint = HMAC_SHA256(IMEI || Iris hash || Battery health)

  • Behavior pattern analysis

- Cursor track entropy value detection

- Analysis of API call time distribution

- Contribution amount mutation alert (>300% triggers manual review)

(10) Economic Attack Protection

  • Redemption flow limit mechanism (the specific value will be set based on the actual operation status of the project)

- Daily redemption limit for single address: 10000 MMI

- Large redemption delays (>5000 MMI required 48 hours to clear)

  • Market Maker Fuse Protection

When the following situations are detected, the redemption will be suspended (the specific value will be set according to the actual operating status of the project):

- MMI price fluctuates >25% within 1 hour

- Fund pool reserve rate <safety threshold (60%)

(11)System disaster recovery plan

  • Multi-chain interoperability

- Main chain: EVM compatible chain or L2 public chain (such as Arbitrum)

- Backup chain: Cosmos application chain

- Cross-chain state synchronization is achieved through LayerZero

  • Emergency control switch

Vault multi-signed address (5/9 signature) can temporarily freeze abnormal accounts

3.2.4 Application layer: Mind Mosaic ecosystem

(1)Web3 task incentive system

The Web3 task system is a major innovation in the Mind Mosaic Identity ecosystem. By providing a rich task selection and reward mechanism, it not only promotes the growth of the community and the marketing promotion activities of Web3, but also brings new business opportunities to developers and eco-participants.

The Mind Mosaic Identity task system is built on a multi-chain compatible smart contract framework. It adopts a modular design to realize the full process automation of "task release-behavior verification-value allocation", forming a task distribution center in the Web3 world.

(2)AI digital human engine

An important part of the AI ​​digital life ecosystem provides users with diverse ways of participating through digital human development and Web3 task system. The system's core logic revolves around the output and redemption of MP points, aiming to motivate users to receive rewards by interacting with AI and contributing data.

(3)SBT(Soul Bound Token)

Cosmic Cipher (CCipher) provides each user with a unique, non-transferable “Mind Mosaic Identity SBT” as the user’s digital identity in the Mind Mosaic ecosystem. This SBT is bound to the user's mobile phone and is used for future scenarios such as airdrop issuance and qualification confirmation.

The AI ​​digital person on the phone is automatically associated with the user's "Mind Mosaic Identity SBT", ensuring that the digital person belongs to the user uniquely. When the digital man exports in GLB format, in addition to including the appearance model, an encrypted SBT reference is embedded.

(4)Cryptocurrency wallet

The wallet is a decentralized, non-custodial digital asset management tool, where users directly manage their assets through private keys or mnemonics. The wallet itself does not hold or store any asset data of the user, and all control belongs to the user.

  • Core functions and technical characteristics

① Decentralized and unmanaged design

Asset Management: The user's private key and mnemonic words are stored locally and kept by the user themselves. The wallet does not have access or control capabilities.

Privacy protection: The wallet does not collect or upload sensitive information from users, and all data operations are completed on the user's device.

② Local identity authentication mechanism

Access method: Unlock access to private keys or mnemonics through device local authentication methods (such as passwords, PINs, fingerprints, or facial recognition).

Security Design: The verification process is performed entirely within the user device, and the private key and mnemonics do not leave the local environment.

③ End-to-end encrypted communication

Data transmission: All communications between the wallet and the blockchain node are end-to-end encryption to ensure that the data cannot be intercepted or interpreted by third parties during transmission.

Protection scope: Only the sender and the receiver can decrypt the data to prevent external eavesdropping or tampering.

④  Blockchain encryption standard

Wallets follow the following industry standards:

a. ECC elliptic curve cryptography: used to generate key pairs and perform digital signatures, providing high-strength encryption.

b. Hash function: used to verify data integrity and transaction consistency.

c. Address derivation: generate wallet addresses according to each network specification.

d. Transaction signature: Ensure the legality and security of the transaction.

e. Compatibility: Through these standards, wallets can interact seamlessly with mainstream blockchain networks.

⑤  Supported blockchain networks

Currently supported: native tokens for networks including BTC, ERC20, Polygon, Solana, Base and other networks.

Expansion plan: Support for more networks will be gradually increased based on technological development and user needs.

  • Technical architecture and security measures

① Architecture basics:

The wallet is centered on decentralization, combining local storage and encryption technologies, and users directly interact with the blockchain network through devices.

② Safety guarantee:

Private keys and mnemonics are stored only on user devices, and the wallet does not involve any cloud backup or remote storage.

Local data encryption storage, external communications are encrypted throughout.

Authentication relies on the device local mechanism and does not require dependence on third parties.

Last updated