Afterlife

Credentials that outlive their owners.

Afterlife is a ghost-access auditor. It pulls identities and credentials from your cloud, code, IdP, and SaaS systems into one identity graph, then runs detection rules over that graph and ranks findings by blast radius.

The class of credentials it surfaces drives a disproportionate share of modern breaches: every API key generated by a contractor who left two years ago, every OAuth grant tied to a deprovisioned employee, every long-lived AWS access key on an account marked "suspended" in the IdP. Uber 2022, Okta 2023, MOVEit, Snowflake 2024 all involve some version of this.

Most tools treat each system in isolation. Afterlife's value is the cross-source identity graph: an AWS access key's "owner" might be active in AWS but suspended in Google Workspace; without joining the two views, you miss it.

Status

v0.3. 9 source systems collected, 16 detection rules, 380+ tests.

Source systems

Plus a CloudTrail enrichment collector (afterlife scan cloudtrail): it adds no identities, but reads recent CloudTrail events to attach observed last-use and used-services to the AWS credentials, giving the usage-based rules ground-truth data. Run it after scan aws.

Each collector lives in src/afterlife/collectors/. They are intentionally dumb (no analysis), idempotent (re-run safe), and tested against mocked APIs (no live calls in CI). Vault's aliases field is used to add cross-system graph edges directly, even without shared email.

What it detects

Each rule's logic, false-positive notes, and remediation are in docs/DETECTIONS.md.

Quickstart

Zero-config demo against in-memory mocks for eight of the nine source systems (all but Okta):

make install
make demo

The demo plants synthetic users, credentials, and IdP records across every collector, runs them, and produces 20 deterministic findings with one OFFBOARDED-OWNER (bob, broad blast) and one ADMIN-CONCENTRATION (dave is admin in 3 systems). Identity graph: 16 persons across 8 sources, 6 cross-source. Demo also writes .afterlife-demo-report.html you can open in a browser.

Against real systems:

make install
.venv/bin/afterlife init
.venv/bin/afterlife scan aws --profile my-profile
.venv/bin/afterlife scan cloudtrail --profile my-profile   # enrich AWS usage
.venv/bin/afterlife scan gcp --project my-project
.venv/bin/afterlife scan github --org my-org --token $GITHUB_TOKEN
.venv/bin/afterlife scan gitlab --group my-group --token $GITLAB_TOKEN
.venv/bin/afterlife scan idp --provider google     # or okta / azure
.venv/bin/afterlife scan slack --token $SLACK_TOKEN
.venv/bin/afterlife scan vault --api-url https://vault.example.com:8200
.venv/bin/afterlife analyze --allowlist allowlist.yaml
.venv/bin/afterlife identities
.venv/bin/afterlife report --format html -o report.html
.venv/bin/afterlife serve                          # localhost dashboard

See .env.example for required environment variables.

Reports

afterlife report emits four formats. Each carries the same finding set; choose based on consumer.

.venv/bin/afterlife report --format pdf -o audit.pdf

Web dashboard

afterlife serve launches a local FastAPI dashboard with nine pages.

More screenshots

Credentials across all sources	Credentials, continued

Findings (medium / low tiers, orphaned identities)	Overview, bottom (top findings by blast radius)

Identities, mixed-source middle	Identities, single-source bottom (bots, no-email cases)

Every scan run captured by the operational scan-history page:

The pages:

• Overview: severity tiles, blast-tier chart, last-scan-per-source.
• Findings: filterable, searchable, sortable, expandable evidence + remediation, one-click ack per finding (state in localStorage), HTMX-powered live filtering.
• Trends: finding history over time from the lifecycle timestamps: open findings by severity, new-vs-resolved flow, and headline stats (open now, seen ever, resolved, median days to resolve).
• Credentials: sortable table with source / type / active filters, click into per-credential detail.
• Identities: person-grouped, filterable to cross-source only, click into per-person detail showing all linked identities + owned credentials + active findings.
• Finding / Credential / Person detail pages: deep-linked, fully cross-referenced.
• Scan history: every afterlife scan ... run with start/end/duration/status.

The dashboard is read-only: no DB writes, no auth, hardened with a strict CSP, X-Frame-Options: DENY, X-Content-Type-Options: nosniff, Cross-Origin-Opener-Policy: same-origin, disabled OpenAPI/docs endpoints, and self-hosted HTMX (no CDN). Dark mode follows prefers-color-scheme. Keyboard shortcuts (/ to search, g h/f/c/i to navigate, ? for help). Includes a print stylesheet for PDF-via-browser.

Alerting

Afterlife tracks findings across runs (each analyze reports what is new, reopened, or resolved since the last run), so it can alert you the moment new ghost access appears rather than only when you go looking.

export AFTERLIFE_SLACK_WEBHOOK=https://hooks.slack.com/services/...
.venv/bin/afterlife analyze --notify        # alert on new/reopened findings

Channels (any combination, configured via environment variables, never persisted):

Only new and reopened findings at or above a severity threshold are sent (AFTERLIFE_NOTIFY_MIN_SEVERITY, default high; or --notify-min-severity), and suppressed findings are never alerted. Run it on a schedule (cron, the CI workflow below) to turn Afterlife into a continuous monitor. See .env.example for every variable.

Continuous monitoring

afterlife run executes the whole pipeline in one pass (scan the configured sources, analyze, and with --notify alert), and afterlife watch repeats it on an interval. Both self-initialize the database, read credentials from the environment, and skip any source that is not configured, so one broken credential never stops the monitor.

# one pass over whatever the environment has credentials for
.venv/bin/afterlife run --notify

# or run continuously, hourly, over an explicit source list
.venv/bin/afterlife watch --interval 3600 --notify -s aws -s github -s idp

Keep the source list and cadence declarative with a config file (see afterlife.example.yml):

.venv/bin/afterlife watch --config afterlife.yml

Docker

docker build -t afterlife .
docker run --rm -v afterlife-data:/data --env-file .env \
    afterlife watch --notify -s aws -s github

The image runs as a non-root user and keeps its database on the /data volume; see the Dockerfile.

CI integration

# .github/workflows/afterlife.yml (excerpt)
- run: afterlife report --format sarif -o afterlife.sarif
- uses: github/codeql-action/upload-sarif@v3
  with:
    sarif_file: afterlife.sarif

Full workflow at .github/workflows/afterlife.yml. It assumes an AWS role via OIDC, scans every source the team uses, uploads SARIF to Code Scanning, and saves an HTML report as a 30-day artifact.

Allowlist / suppression

afterlife analyze --allowlist allowlist.yaml reads a YAML file naming findings to suppress. Suppressed findings are persisted (still auditable) but hidden from the default dashboard view. Example:

- rule_id: NEVER-USED
  credential_id: arn:aws:iam::123:role/SeasonalReportingRole
  reason: Yearly audit role, intentionally dormant
  until: 2027-01-01

Matchers: rule_id, credential_id, identity_source, identity_id. All named fields must match. Catch-all entries (no matchers) are refused at load time.

Editions

Afterlife is open core. Everything above is free and always will be: detection, the identity graph, blast scoring, monitoring with history, alerting, run / watch, reports, and the local dashboard.

Pro adds team/enterprise features, unlocked by an offline license key (a signed token verified locally against an embedded public key -- no license server, nothing phones home):

Get Pro (from $990/year, founding rate, first 50 organizations, price locked for life): email didesle7@gmail.com with your organization name and you'll get a signed license key to activate:

afterlife license                       # show your edition
export AFTERLIFE_LICENSE=<token>        # activate (or AFTERLIFE_LICENSE_FILE=<path>)

_{Maintainers mint licenses with scripts/issue_license.py using the vendor private key, which is never committed.}

Architecture

  collectors/  ─►  SQLite  ─►  identity graph  ─►  rules engine  ─►  blast scoring  ─►  reports
   AWS / GCP                   (NetworkX,           (pluggable)        (per-finding)     (json/html
   GitHub / GitLab              email + Vault                                             /sarif/pdf)
   Google / Okta / Azure        aliases)
   Slack / Vault

Five layers, each with a narrow boundary. Collectors write to SQLite only. Rules read from SQLite + identity graph only. Scoring is pure (input: credential, output: blast radius). Reports are pure (input: DB, output: text). The dashboard wraps the same readers behind FastAPI. Details in docs/ARCHITECTURE.md.

Project layout

src/afterlife/
├── collectors/    9 source collectors + a CloudTrail usage-enrichment pass
├── rules/         16 detection rules, one file each, decorator-registered
├── graph/         Identity graph (NetworkX), email + Vault-alias linking
├── scoring/       Blast-radius scoring with explainable factors
├── reporting/     JSON, HTML, SARIF, PDF
├── notify/        Alerting: Slack, webhook, email (SMTP)
├── web/           FastAPI dashboard + templates + static assets
├── runner.py      Pipeline: scan -> analyze -> notify (run / watch)
├── allowlist.py   YAML suppression loader + matcher
├── licensing.py   Offline Pro-license verification (open core)
├── scan_runs.py   Run-tracking context manager
├── db.py          SQLite schema + helpers
├── models.py      Identity, Credential, Finding, BlastRadius
└── cli.py         Typer CLI

tests/             380+ tests using moto, respx, freezegun, fastapi.testclient
demo/              Self-contained `make demo` (mocks for every collector)
docs/              ARCHITECTURE.md, DETECTIONS.md, INTERVIEW_TALK_TRACK.md
.github/workflows  Production-ready GitHub Action

Further reading

• docs/PLAYBOOK.md: complete usage walkthrough (start here for hands-on)
• docs/ARCHITECTURE.md: layered design + why these boundaries
• docs/DETECTIONS.md: every rule, false-positive notes, remediation
• docs/blog/the-graph-layer.md: design essay on why a graph is the right shape for cross-source ghost-access detection
• docs/INTERVIEW_TALK_TRACK.md: prepared narratives for portfolio conversations
• docs/KEY-MANAGEMENT.md: backing up and rotating the license-signing key (maintainers)
• CHANGELOG.md: full release + milestone history

Why "Afterlife"

The credentials this tool finds shouldn't still be alive.

License

MIT