Decisions in changing systems.
Research Problem
Cloud-native distributed systems make access decisions while identities, request context and operational conditions change. Roles and attributes provide explicit policy boundaries, but understanding how authorization should respond to abnormal behavior remains a research challenge.
AeglysAI investigates whether behavioral evidence and operational system state can inform useful adaptations while preserving deterministic constraints and understandable decisions. The benefits and tradeoffs remain to be tested.
How does explainable, policy-bounded, closed-loop adaptive authorization perform relative to RBAC, ABAC and deterministic risk-aware authorization under behavioral and operational anomalies in cloud-native distributed systems?
Baselines
Implementation status refers to the separate AeglysAI research/application repository. This portal presents the research program.
RBAC
Role-based access control establishes the deterministic authorization baseline.
ImplementedRBAC + ABAC
Attribute-based conditions extend role-based authorization with request context.
ImplementedContextual Risk
Deterministic contextual risk adds explicit risk rules to role-based authorization.
ImplementedModel D — Research Implementation Complete
The bounded offline research implementation is complete; internal synthetic validation has been performed. Model D remains experimental and non-enforcing.
Completed work includes offline behavioral inference, eligibility and drift diagnostics, experimental integration, mock-feedback simulation and internal reproducibility checks. These research capabilities do not grant or deny access.
enforcement_authority=NONE. Internal synthetic validation does not establish external field effectiveness, production authorization capability or commercial readiness. Drift coverage and mock-feedback effectiveness limitations remain unresolved.
Local engineering acceptance is not independent field validation, security/distribution clearance or a commercial product release. No model artifacts, datasets or internal evaluation results are published here.
Research Principles
Policy remains explicit
Investigate adaptation constrained by deterministic rules and reviewable decision boundaries.
Evidence before claims
Separate implemented capabilities, planned designs and experimentally supported findings.
Explain the decision
Make the contributing policy, context and proposed risk evidence understandable to reviewers.
Reproducibility
The intended experimental program compares models under documented scenarios and controlled conditions. Methods, configurations and limitations should accompany results so others can assess and reproduce the work.
View the experimental plan →Research hypotheses
Research hypotheses will be published after the research specification is finalized.