Table of Contents

Security

The Security section examines architectural boundaries that can reduce accidental authority, hidden execution paths, unsafe defaults, and ambiguous control flow.

Security in ASI Backbone Learning is approached as an architectural responsibility rather than a single feature or package.

Section status: Focused security learning now covers trust boundaries, least privilege, replay protection, and cryptographic evidence boundaries. Start with Trust Boundaries and Least Privilege, continue with Replay Protection and Bounded-Use Authority, then study Signing, Verification, Key Custody, and Tamper Evidence before using the Foundational Tutorials to connect security boundaries to governed execution.

A secure boundary should remain visible when the system is under pressure.

Start Here

Trust Boundaries and Least Privilege is the first focused security tutorial. It treats trust boundaries as changes in control over data or authority and least privilege as an architectural constraint on what authority crosses those boundaries.

The tutorial connects caller-supplied versus authoritative context, authentication, authorization, policy decisions, credential ownership, narrow authority, boundary validation, resource ownership, and fail-safe behavior.

Replay Protection and Bounded-Use Authority continues from narrow authority into stateful execution-boundary enforcement. It covers one-time and bounded-use grants, atomic consumption, multi-instance and restart behavior, durable replay state, failure windows, request idempotency, and why replay resistance is not an exactly-once execution guarantee.

Run the Replay Protection and Bounded-Use Authority sample to compare a deterministic check-then-act race with atomic in-process consumption, then use the intermediate concurrency lab to repair and extend the boundary.

Signing, Verification, Key Custody, and Tamper Evidence adds the cryptographic evidence boundary. It separates hashes from signatures, signing from verification and authorization, key custody from ordinary application configuration, normal rotation from compromise, and tamper evidence from tamper prevention.

Security Themes

Current and future material may examine:

Approval Is Not Unlimited Authority

A recurring security principle in the foundational material is:

Allowed Decision
   ≠
Broad Standing Permission

An operation that has been approved may still benefit from authority that is:

  • Narrow
  • Bound to a specific actor
  • Bound to a specific operation
  • Bound to a specific resource
  • Bound to an intended executor or audience
  • Time-limited
  • Revalidated before execution

This concept is explored in:

Scoped Capability and Host-Owned Execution

Security and AI-Assisted Systems

For AI-assisted workflows, prompt instructions are not treated as security controls.

The central boundary remains:

The model may propose. The host retains execution authority.

Host-side code remains responsible for:

  • Tool allowlists
  • Argument validation
  • Authoritative context
  • Policy evaluation
  • Secret ownership
  • Destination and egress controls
  • Capability validation
  • Real-world execution

See:

Governed AI Tool Gateway

Working References

Security-related implementation examples can be studied in:

The first focuses on governed decision and execution boundaries.

The second provides a broader ASP.NET Core reference architecture with secure application defaults and operational controls.

Scope

The material in this section is educational.

It does not constitute:

  • A security certification
  • A penetration test
  • A formal threat model
  • A compliance assessment
  • A guarantee that a demonstrated pattern is sufficient for production

Application-specific security analysis remains necessary.

Current Status

The Security section now has focused tutorials for trust boundaries and least privilege, replay protection and bounded-use authority, and signing/verification with key-custody and tamper-evidence boundaries. Together they establish a security architecture path from identifying where trust changes, to preserving narrow authority, to controlling whether authority may be consumed again, to deciding what cryptographic evidence can safely establish across trust boundaries.

Future material will extend into capability-based authority, secure logging, dependency integrity, and threat modeling.

Use the Foundational Tutorials to connect these security concepts to the existing governed-execution learning path.

Read it. Run it. Question it. Improve it.

Edit this page