Data Chain of Custody

How ShieldNex is designed to work

How an image is intended to become something you can act on: recorded sources, integrity checks, a quality gate that can return no answer at all, and a human reviewer between the model and you.

Six steps, two of which can stop the process

The quality gate and the human reviewer both exist to prevent a weak observation reaching you. This describes the intended design of the platform, which is still in development.

Recorded sources

Open-data baseline & licensed commercial imagery

Imagery is drawn from open government programmes and, where a case warrants it, licensed commercial providers. Source, licence terms and acquisition time are recorded with every frame so that each observation can be traced back to the pass it came from.

Open-data programmes (ESA Sentinel, USGS/NASA Landsat)
Licensed commercial high-resolution imagery where warranted
Source, licence and acquisition time recorded per frame
Evidence generation is blocked if licence metadata is missing

Feeds into next step

Integrity checks on arrival

SHA-256 Cryptographic Hashing

Every image is hashed at the moment of ingestion, so that a later alteration to stored data can be detected by re-checking the hash. Hashing is one provenance control among several, on its own it does not establish authenticity or admissibility.

SHA-256 hash generated at point of ingestion
Hash chain maintained from ingestion to delivery
Alteration of stored data becomes detectable
One control among several, not a guarantee on its own

Feeds into next step

Image quality check

Abstain rather than guess

Before any analysis runs, the pair of images is checked for cloud, haze, shadow, seasonal variation and alignment. If the comparison cannot support a reliable observation, the result is returned as inconclusive. An honest "we cannot tell from this imagery" is a valid output.

Cloud, haze and shadow screening
Seasonal variation and image alignment checks
Explicit inconclusive state for unusable comparisons
No detection is attempted on imagery that fails the gate

Feeds into next step

Change detection

Limited, defined change classes

Temporal comparison flags a deliberately narrow set of observable physical changes: a new structure or extension, land clearing or earthwork, and a new access track or enclosed compound. Walls, fences and small structures are below 3 m resolution and are checked only on 0.5 m imagery. Satellite resolution does not establish cadastral accuracy, so no claim is made about boundaries or title.

Temporal comparison across satellite passes
New structure or extension · clearing or earthwork · new track or compound
Reports observable physical change only
No boundary, cadastral or ownership determination

Feeds into next step

Human review

Mandatory before anything leaves the system

A candidate change is not an alert. A human reviewer must accept it first, and that decision is recorded as part of the observation. Nothing is sent to an owner, and no evidence package is generated, on the model's word alone.

Reviewer decision required on every candidate
Rejected and inconclusive outcomes are retained, not discarded
Reviewer identity and decision recorded with the observation
Alerting only follows an accepted review

Feeds into next step

The evidence package

IT Act 2000 & Bharatiya Sakshya Adhiniyam 2023 aligned

An accepted observation is packaged with the production history behind it: before/after imagery, source and licence metadata, timestamps, processing and model versions, the reviewer's decision, hash values and access history. It is a preliminary input to verification, not proof, and not a claim of admissibility.

Timestamped before/after satellite imagery
Source, licence, processing and model version records
Reviewer decision and access history
Material structured to support a Section 63 (BSA 2023) certificate

Data sources & cadence

Which satellites we use, and how often

The intended approach combines open baseline imagery with licensed commercial tasking, so a plot can be looked at with the resolution a case actually needs. Revisit intervals below are the published characteristics of each mission, not a monitoring cadence we are currently committing to.

Open baseline · routine comparison

Open-data foundation

ESA Sentinel-2
10 m multispectral optical imagery, roughly every 5 days at the equator. Intended primary source for routine change detection.
ESA Sentinel-1 (SAR)
C-band synthetic-aperture radar that sees through cloud cover, revisit every 6–12 days. Intended as the monsoon fallback when optical imagery is clouded out.
USGS / NASA Landsat 8 & 9
30 m multispectral optical, 16-day revisit, with a deep historical archive stretching back over a decade for before-and-after baselining.

Commercial high-resolution · on demand

Sharper imagery, bought when it is needed

3 m commercial imagery
Not licensed yet
The intended core source: next-day 3 m commercial imagery of each enrolled plot. No commercial licence is held today.
0.5 m imagery
Not licensed yet
Bought only when a 3 m check finds something, to confirm it and to see walls, fences and small structures. One capture covers at least 25 km², so it is used to confirm, not to monitor every plot.

We make no claim of any provider relationship before a licence is held.

What we will not promise about cadence

A satellite's revisit interval is not the same thing as a monitoring cadence. Whether a given pass produces a usable comparison depends on cloud, haze, shadow and seasonal conditions, and on what our imagery licences permit. We are not currently in a position to state how often a usable comparison will be available for a given plot, so we do not advertise a scan frequency or an alert window.

We also hold no tasking commitment from any provider. Where commercial procurement is described on this page, it is planned, not secured.

Real-time video surveillance is not offered by any Earth observation system today and is not part of what we provide.

Built on India's geospatial framework

ShieldNex works with satellite imagery under India's Guidelines for Acquiring and Producing Geospatial Data and Geospatial Data Services (2021), which liberalised the use of such data for Indian entities. Where any airborne or drone capture is used, those specific operations are conducted in line with the applicable DGCA drone rules.

The legal framework we design against

Our evidence packages are designed to preserve the production history that the electronic-record procedures under current Indian law rely on. We make no claim that our output is admissible or court-ready — admissibility and evidentiary weight are for the court to determine.

IT Act 2000

Recognises electronic records and gives statutory footing to hash-based integrity checks

Bharatiya Sakshya Adhiniyam 2023

Electronic records under Sections 61–63, supported by a Section 63 certificate (replaced the Evidence Act, 1872 from 1 July 2024)

Geospatial Guidelines 2021

Satellite data used within India's liberalised geospatial data framework

What hashing does, and what it does not

The chain of custody is designed so that an alteration to stored evidence, by any party, ShieldNex included, becomes detectable on re-checking the hash.

That is a narrower claim than it may sound. Hashing shows that a record has not changed since it was hashed. It says nothing about whether the original imagery was correctly sourced, whether the analysis was right, or whether a court will accept the result. Those depend on the licence records, the quality gate, the reviewer's decision and expert testimony — which is why the hash is one control here rather than the whole argument.