AeglysAI Behavioral Risk Challenge

Problem

Behavioral signals can be noisy, and risk assessments need evidence that distinguishes suspicious activity from legitimate variation.

Challenge

Prototype or evaluate behavioral risk methods using synthetic or appropriately licensed data. Explain signal selection, assess false positives and document uncertainty. Model D is planned research, not an existing implemented capability.

Timeline

Planned launch: December 2026
Planned submission: February 2027

Exact deadlines, time zones and review dates will be announced in official terms.

Eligibility

Free participation for platform engineers, cloud engineers, security engineers, SREs, software engineers, AI/ML engineers, researchers and graduate students. Final eligibility, team limits and any geographic or age requirements will be specified in official event terms.

Deliverables

• Prototype or evaluation code with reproducible setup
• Data provenance, signal definitions and privacy considerations
• Evaluation method, false-positive analysis and explainable risk evidence

Proposed scoring framework

• Technical Correctness — 25%
• Security & Threat Reasoning — 20%
• Architecture & Engineering — 20%
• Reproducibility — 15%
• Innovation — 10%
• Documentation & Explainability — 10%

Independent judges score eligible submissions using the same evidence-based rubric. Conflict disclosure, recusal and a published tie procedure protect independent review. The framework remains subject to finalized official terms.

Prize

Planned USD $300 prize. Prize eligibility and payment are subject to official event rules. Prizes and rules are subject to official event terms until finalized. No prize has been awarded. Award conditions and payment details remain to be finalized.

Submission process

Required submission package (subject to finalized official terms):

• Public GitHub repository unless an exception is approved before submission
• README
• Problem statement
• Architecture description
• Setup instructions
• Demo
• Test evidence
• Security considerations
• Limitations
• License information

Optional: Video demo, Benchmark results, Research notes. A demo is required; a video is optional. Submit the repository URL and a specific commit/tag through the official form when available. Receipt is separate from eligibility and judging.

Opening Soon. Submissions are not open yet.

Rules

Proposed rules: use authorized isolated environments; avoid real credentials or personal data; disclose reused code, datasets and AI assistance; credit collaborators; and report results honestly.

Final eligibility, team size, permitted prior work, submission deadlines and award procedures are subject to official event terms.

Open-source expectations

A public GitHub repository is required unless an exception is approved before submission. Include source code, setup instructions, evaluation artifacts and explicit license information. Respect upstream licenses and document third-party dependencies and data provenance.

Use synthetic or appropriately licensed data; never publish secrets or sensitive datasets. Final licensing and submission requirements will be defined in official terms.

Tracks

Event-specific tasks, resources and track evaluation: TBD.

Judges and results

Independent judges: TBD. No appointments or results are claimed.

Results will be published only after eligibility checks, independent judging and organizer approval.

Shared participant guidance

Qualification and GitHub workflow →

Certificate eligibility framework →

Final submission does not guarantee an upstream merge. Maintainers retain normal review standards.