Skip to content

venkatapgummadi/ascend

Repository files navigation

ASCEND

Automated Scanning, Compliance ENforcement, and Deployment

A four-layer DevSecOps framework that integrates automated security scanning directly into CI/CD pipelines with build-gating mechanisms, multi-track deployment orchestration, and AI-powered post-deployment code synchronization.

CI License: MIT Python 3.9+ Paper GitHub stars Last commit PRs welcome


Overview

Modern CI/CD pipelines prioritize velocity. ASCEND's thesis is that velocity and security are not in tension — the tension comes from treating security scanning as a passive reporting step rather than as an active build gate. ASCEND integrates security scanning, build gating, multi-track deployment orchestration, and AI-powered synchronization into a single, platform-agnostic framework.

Four-Layer Architecture

┌─────────────────────────────────────────────────────────────────┐
│  Layer 1: Source Analysis                                        │
│  SAST (CodeQL, Semgrep, SonarQube) + SCA (Snyk) + Secrets       │
│                           [ Quality Gate 1 ]                     │
├─────────────────────────────────────────────────────────────────┤
│  Layer 2: Build & Integration                                    │
│  Container Scan (Trivy) + IaC Scan (Checkov) + License Check    │
│                           [ Quality Gate 2 ]                     │
├─────────────────────────────────────────────────────────────────┤
│  Layer 3: Deployment Orchestration                               │
│  Blue-Green / Canary / Rolling + DAST (OWASP ZAP)               │
│                           [ Quality Gate 3 ]                     │
├─────────────────────────────────────────────────────────────────┤
│  Layer 4: AI-Powered Synchronization                             │
│  AST Drift Detection + ML Conflict Classification + LLM Resolve  │
│                      [ Back-propagation ]                        │
└─────────────────────────────────────────────────────────────────┘

Key Contributions

  1. Four-layer DevSecOps architecture with formal quality gate definitions.
  2. Platform reference configurations for GitHub Actions, GitLab CI/CD, Jenkins, and Azure DevOps.
  3. Multi-track deployment framework supporting blue-green, canary, and rolling strategies with automated quality gates at each promotion boundary.
  4. AI-powered synchronization system using AST differencing, ML conflict classification, and LLM-based resolution with property-based verification.
  5. Comprehensive scanning tool integration covering SonarQube, Semgrep, CodeQL, Snyk, Trivy, OWASP ZAP, Checkov, and TruffleHog.

Quick Start

1. Choose your platform

ASCEND provides full reference configurations for four major CI/CD platforms:

Platform Location Best For
GitHub Actions platforms/github-actions/ Teams on GitHub with GHAS
GitLab CI/CD platforms/gitlab-ci/ Fastest integration via native templates
Jenkins platforms/jenkins/ Existing Jenkins infrastructure
Azure DevOps platforms/azure-devops/ Microsoft enterprise ecosystems

2. Copy the pipeline configuration

Example for GitHub Actions:

cp platforms/github-actions/.github/workflows/ascend-full.yml \
   /path/to/your/repo/.github/workflows/

3. Configure scanning tools

Each scanning tool requires minimal configuration (API tokens, organization IDs). See quality-gates/README.md for setup instructions for each tool.

4. Enable the AI synchronization layer (optional)

cd ai-sync
pip install -e .
ascend-sync --help

See ai-sync/README.md for configuration.

5. Run your first pipeline

Push a commit to your feature branch. ASCEND will execute Layer 1 scanning immediately. Passing builds progress through Layer 2, Layer 3, and Layer 4 according to your promotion rules.


Reproducing the results

ASCEND's evaluation is designed to be reproducible from public inputs only. The harness in benchmark/ regenerates every number reported in the paper; there is no private or NDA-gated data in the loop.

cd benchmark
pip install -r requirements.txt

# 1) Detection coverage on the labeled OWASP Benchmark (real TP rates)
python run_coverage.py --benchmark ./_data/owasp-benchmark \
                       --sarif-dir ./_data/sarif --out results/coverage.json

# 2) Conflict-resolution metrics on the public conflict fixtures
python run_sync.py --fixtures ../examples/conflict-fixtures \
                   --tau 0.85 --bootstrap 10000 --out results/sync.json

# 3) Per-layer scanning overhead on the sample apps
python run_overhead.py --app ../examples/sample-python-app \
                       --runs 5 --out results/overhead.json

# 4) Emit LaTeX rows for the paper tables
python gen_latex.py results/*.json

What the harness measures and how to read it:

  • Detection coverage is computed against the OWASP Benchmark's labeled ground truth, so single-tool vs. multi-tool true-positive rates and the marginal value of each added tool are directly verifiable.
  • Conflict resolution is reported as an auto-resolve rate and a conditional accept-rate — the latter applies only to the subset the system chooses to auto-resolve, and is not a general accuracy figure. Accept-rate comes with a 95% bootstrap confidence interval and a per-type breakdown.
  • Overhead is wall-clock time per scanning layer (median and P95); only installed tools are timed.

We deliberately do not claim a multi-organization field study. Earlier drafts reported pre/post metrics from private repositories; those claims are withdrawn because they cannot be independently reproduced, and the project now reports only public-benchmark results.


Adoption Roadmap

ASCEND is designed for incremental adoption. Most organizations realize the majority of security value from Layer 1 alone.

Phase Effort Layers Outcome
Phase 1 1–2 weeks Layer 1 ~80% of vulnerability reduction
Phase 2 4–6 weeks Layers 1–2 Container & IaC coverage
Phase 3 4–6 weeks Layers 1–3 Multi-track deployment gates
Phase 4 8–12 weeks Layers 1–4 Full AI synchronization

Start with Phase 1, measure impact, and progress only when the current phase is operating smoothly.


Repository Structure

ASCEND/
├── docs/                    # Architecture docs and research paper
│   ├── architecture.md
│   ├── quality-gates.md
│   ├── adoption-guide.md
│   └── paper/               # Research paper sources and metadata
├── platforms/               # Platform-specific CI/CD configurations
│   ├── github-actions/
│   ├── gitlab-ci/
│   ├── jenkins/
│   └── azure-devops/
├── ai-sync/                 # AI synchronization Python module
│   ├── ascend_sync/         # Source package
│   ├── tests/
│   └── examples/
├── quality-gates/           # Scanning tool configurations
│   ├── sonarqube-quality-gate.json
│   ├── semgrep-rules.yml
│   ├── checkov-config.yml
│   ├── zap-rules.tsv
│   └── trufflehog-config.yml
├── examples/                # Sample applications with ASCEND integrated
└── scripts/                 # Setup and validation utilities

Documentation

Getting started

Reference

Enterprise and compliance

Project information

Examples

Working sample applications with ASCEND pre-integrated — see examples/ for details.


Research paper

ASCEND is described in a working manuscript that presents the four-layer architecture, the composite quality-gate rule, and the verification-gated synchronization loop, evaluated on the public benchmark above.

Status: preprint / under submission. Not yet peer-reviewed or accepted. Cite the archived software release (below) for reproducibility.

Citation:

@misc{gummadi2026ascend,
  title  = {ASCEND: A Verification-Gated DevSecOps Framework with
            AI-Assisted Synchronization},
  author = {Gummadi, Venkata Pavan Kumar},
  year   = {2026},
  note   = {Preprint; under submission. Evaluated on a public benchmark.}
}

See CITATION.cff for additional citation formats.


Contributing

Contributions are welcome. Please read CONTRIBUTING.md for the contribution workflow and CODE_OF_CONDUCT.md for community standards.

Areas where contributions are especially valuable:

  • Additional platform configurations (CircleCI, TeamCity, Bamboo, Buildkite)
  • Additional scanning tool integrations
  • Conflict resolution model training data (anonymized merge conflict histories)
  • Language-specific SAST rule sets
  • Sample application integrations

License

ASCEND is released under the MIT License. See LICENSE for the full license text.


Author

Venkata Pavan Kumar Gummadi IEEE Senior Member | Professional Software Engineer Email: venkata.p.gummadi@ieee.org


Acknowledgments

The ASCEND framework draws on the public DevSecOps knowledge base assembled by the broader security engineering community, including the NIST Cybersecurity Framework (CSF 2.0), the NIST Secure Software Development Framework (SSDF, SP 800-218), CIS Benchmarks, and the authors of the scanning tools integrated into this framework.

About

DevSecOps for the AI-era CI/CD pipeline. Catches the bugs your AI coding assistant introduces — before they reach production.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

20 stars

Watchers

0 watching

Forks

Packages

 
 
 

Contributors