Architecture & Security
'Under the Hood' details, encryption mechanisms, and privacy levels.
Understanding data processing protocols is critical for ensuring security and privacy. The deAPI architecture utilizes a decentralized network of verified Workers, maintaining a balance between high-performance inference and robust data protection.
1. Under the Hood: Security Architecture
The following section outlines the data flow for a standard inference job (using Image-to-Image as the reference model).
Data Flow
Client → deAPI Server (HTTPS Encrypted) Requests are initiated via the API. All data transmission is encrypted in transit using HTTPS protocols.
Server → Worker The central server designates an appropriate Worker node and transmits the model, parameters, and task specifications.
Worker (Processing) The Worker node receives the payload (image, prompt, settings). Data is processed by a local, isolated Python server.
Result → Client The generated or modified asset returns to the client via the same encrypted return path.
Security Mechanisms
Multi-layer security measures are implemented to guarantee network integrity and data protection:
HTTPS Everywhere: Encryption is enforced at every stage of communication.
Request Hashing: Checksums are utilized to prevent Worker spoofing or the submission of fraudulent results.
Binary Verification: The Worker application automatically validates checksums for the Python server and models prior to execution, ensuring code integrity.
Injection Protection: The execution environment is strictly isolated to prevent code injection that could compromise Worker infrastructure.
2. Privacy Model & Worker Reliability
Security in the distributed environment is maintained through a reputation system and strict technical data access controls.
Data Visibility
deAPI Server: Accesses prompts and metadata required for job coordination. No long-term storage of sensitive payload data occurs.
Worker: Receives the full payload (e.g., input image), but data resides exclusively in RAM. No disk writes occur, and there is no exposure via a UI. Data extraction would require advanced, real-time reverse engineering.
Additional Safeguards
Worker Scoring: An automated reputation system identifies and excludes unreliable nodes that fail to complete tasks.
Dual Verification (Optional): For sensitive workloads, task verification by two independent Workers is available (incurring 2x cost, while remaining significantly below standard cloud pricing).
3. Use Cases: Decentralized vs. Enterprise
The decentralized model is classified as a "Perfect Fit" for: ✓ Public content transcription (YouTube, X, TikTok) ✓ Image and video generation applications ✓ AI workloads devoid of critical data (PII, trade secrets)
For these scenarios, Community Workers provide an optimal price-to-performance ratio (estimated 5-10x cost reduction compared to centralized cloud), without unnecessary expenditure on Enterprise-grade infrastructure.
*Secure Worker Tier For Enterprise clients handling sensitive data (internal documentation, prototypes), a Premium tier is being introduced featuring:
Dedicated GPU farms (RTX 4090, H100, RTX 6000 PRO) located in secure Data Centers.
Full infrastructure isolation, audited operators, NDAs, and legal accountability.
Certified privacy standards (at a higher price point)
*Additional request - contact support for more information: [email protected]
Last updated