Sunday, October 18, 2026

EVS Clearance Levels & Operational Mechanics

 





Does your System Require Strict Regulatory Compliance: Military, Utility, Justice Department or Healthcare?


Clearance Levels & Operational Mechanics: EVS is available Free for Power11 Users with Free Deployment Support:




_______________________________________________


Equitus Video Sentinel (EVS)  is deployed alongside IBM Federal  and supports U.S. Customs and Border Protection (CBP) and Department of Defense (DoD) programs, it operates across multiple levels of security clearances and facility authorizations.


1. Levels of Security Clearances Supported

Equitus–IBM deployment operates across three main operational tiers:

A. CUI / Public Trust / Secret (Border & Tactical Edge Operations)

  • Public Trust (High Risk) & Secret Clearance: Field deployments supporting CBP, Department of Homeland Security (DHS), and tactical border monitoring utilize personnel and systems cleared at Secret or High-Risk Public Trust (e.g., SF-86 / SF-85P background vetting).

  • Controlled Unclassified Information (CUI / DoD Impact Level 4 & 5): Handles sensitive operational video feeds, cross-border sensor networks, and real-time tracking metadata that are unclassified but restricted under federal compliance frameworks.

B. Top Secret / SCI (Defense & Intelligence Installations)

  • TS/SCI (Top Secret / Sensitive Compartmented Information): For deep intelligence fusion, military command centers, and Joint All-Domain Command and Control (JADC2) programs, Equitus EVS integrates into IC-IG (Intelligence Community) networks.

  • Cleared Personnel & SCIF Environments: IBM Federal engineers and Equitus cleared field personnel holding TS/SCI with Polygraph maintain and configure software pipelines operating inside Sensitive Compartmented Information Facilities (SCIFs).

C. Air-Gapped / Sovereign Zero-Trust Deployments

  • IL6 / Special Access Programs (SAP): Equitus EVS is designed to operate in completely air-gapped environments. It requires zero outbound cloud connections to process video, meaning classified sensor data remains fully contained within localized military enclave boundaries.





2. How the Platform Operates at These Clearance Levels






1. Edge Inference Without Cloud Egress

In high-security environments, video streams cannot be sent to commercial public clouds due to data sovereignty and bandwidth constraints. Equitus EVS runs vision models directly on IBM Power10/11 on-prem hardware at the local edge station. Because processing happens locally on-chip, raw surveillance video never leaves the air-gapped perimeter.

2. Secure Metadata Correlation via KGNN

Instead of transmitting heavy, unencrypted video feeds across networks, EVS extracts structured metadata (e.g., timestamps, spatial coordinates, bounding boxes, vector embeddings). This metadata is fed into the Equitus Knowledge Graph Neural Network (KGNN) running within the classified enclave, bridging legacy intelligence databases with real-time border telemetry.

3. Red Hat Enterprise Linux (RHEL) & OpenShift Security Controls

Equitus EVS is tuned for RHEL and Red Hat OpenShift running on IBM infrastructure. This allows IBM Federal to enforce FIPS 140-3 cryptography, SELinux mandatory access controls, and DISA STIG compliance out of the box, fulfilling federal requirements for both classified networks (SIPRNet/JWICS) and unclassified federal networks (NIPRNet).

________________


Complimentary Commercial Open Source Models: used by developers and researchers to help computers "see" and "understand" visual data through different specialized tasks: Allows Power11 Users to integrate with or without Cloud.  IBM EVS Power 11 systems can communicate utilizing both MMA and GPU micro-processing.


  • YOLO (You Only Look Once): Primarily used for real-time object detection, which means it identifies and draws bounding boxes around objects in images or videos.
  • SAM (Segment Anything Model): Developed by Meta, this is a foundation model used for image segmentation, allowing users to cut out or isolate any specific object within an image with a single click.
  • RelateAnything: An extension built on top of models like SAM, used for relationship detection or scene graph generation, which helps identify how different objects in an image interact with or relate to one another.


No comments:

Post a Comment

Equitus Video Sentinel (EVS)

  "Zero-Trust Visual AI" Proof-of-Concept (PoC) IBM EVS - Advanced multi-modal video analytics stack—featuring IBM EVS/Video Senti...