Data Validation

Our current standpoint is to rely on two main validation principles: validation based on the challenge-response concept (where validation of every single action is unfeasible) and direct proof mechanisms (where validation of every single action is feasible). The following section introduces the validation mechanisms available.

Proof of Location

While every Detector can be uniquely identified, it is not sufficient to rely on location data provided by Detectors; this is due to the ease with which GPS information can be spoofed. NATIX Network’s key feature is the location specificity of the data. We rely on proof-of-location mechanisms to ensure that the data actually comes from a claimed location. We plan to leverage decentralized location attestation solutions such as XYO network to ensure even greater accuracy as well.

Proof of Live Stream

Proving a Detector’s node is analyzing a real-time video stream, not a pre-recorded video, is an important part of data validation. The primary mechanism used by the protocol relies on the challenge-response concept. Here, a randomly assigned xNode will challenge the Detector to execute a randomly generated set of actions on camera. The exact algorithm for the challenge-response mechanic will be released in the technical whitepaper, including the slashing conditions in case of equivocation.

Proof of Computation

The protocol ensures that the real-time geospatial data is correctly extracted by machine learning predictions (AI) provided in the mApplet and not generated heuristically.
Zero-knowledge proof schemes address this issue. Solutions like zkCNNs, vCNNs, and pvCNNs are zero-knowledge proof for convolutional neural networks. Some of these approaches ascertain model secrecy. Such schemes allow the owner of the AI model to provide proof without leaking any information about the model itself. They also provide proof that AI produced the output.
NATIX Vision SDK will incorporate verifiable AI compute mechanisms when training the AI models. This will ease the process for Developers to create mApplets that are capable of providing such proofs.
There are multiple other validations and proof mechanisms that NATIX is currently working on and are in the R&D phase. We will provide more information on these in later stages.