The Problem

Thousands of CVEs are published every month. As a solo security researcher doing bug bounty work and vulnerability assessments, I needed a way to cut through the noise and surface the vulnerabilities that actually matter — the ones being actively exploited, the ones with public proof-of-concept code, and the ones most likely to show up in the environments I test.

Commercial vulnerability intelligence platforms exist, but they're expensive and opaque about how they prioritize. I wanted something I controlled, understood end to end, and could tune to my own workflow.

What It Does

The CVE Intelligence Dashboard is a self-hosted pipeline that continuously ingests vulnerability data from multiple public sources, normalizes it into a single database, scores each CVE by real-world risk, and serves the results through a web dashboard and Telegram bot.

It only aggregates publicly available vulnerability data. It performs no scanning, testing, or probing of any third-party system.

Data Sources

The ingestion pipeline pulls from six public vulnerability data feeds:

Source What It Provides
NVD (NIST) CVE descriptions, CVSS scores and vectors, CWE classifications, CPE (affected software), and references
MITRE CVE records including those not yet analyzed by NVD
CISA KEV Known Exploited Vulnerabilities catalog — confirmed active exploitation, vendor/product data, ransomware association, and remediation deadlines
EPSS Exploit Prediction Scoring System — probability that a CVE will be exploited in the wild within 30 days, with historical score tracking
OSV / GHSA Open Source Vulnerabilities and GitHub Security Advisories — open-source package-level vulnerability data
VulnCheck Additional enrichment data (optional, API-key gated)

Each source has its own ingester module. Ingestion runs on a cron schedule with file-level locking to prevent overlapping runs. The pipeline tracks ingestion state per source so it can resume cleanly after failures.

Priority Tiering

Raw CVSS scores alone are a poor signal for real-world risk. The dashboard combines multiple factors into a priority tier that reflects actual exploitability:

  • ACT_NOW — in the CISA KEV catalog (confirmed exploitation), associated with ransomware campaigns, or the combined CVSS × EPSS priority score exceeds the threshold
  • HIGH — elevated priority score, or critical/high severity with significant EPSS probability
  • MEDIUM — medium or high severity CVEs
  • INFO — everything else worth tracking

The system also computes a bug bounty relevance score (0–100) based on attack vector, authentication requirements, complexity, user interaction, CVSS severity, EPSS probability, and PoC availability. This powers the hunt list — a filtered, ranked view of CVEs most relevant to web application testing.

Dashboard Features

The Flask web dashboard provides:

  • Smart search with structured query parsing — filter by vendor:, product:, cwe:, severity:, EPSS/CVSS thresholds, KEV membership, PoC availability, and no-auth requirements
  • Browse — paginated view of all tracked CVEs with multi-filter controls
  • KEV catalog — CISA Known Exploited Vulnerabilities with vendor filtering
  • Critical CVEs — critical-severity CVEs ranked by EPSS exploitation probability
  • Trending — CVEs with the highest current EPSS scores
  • Ransomware — KEV entries flagged for known ransomware association
  • PoC tracker — proof-of-concept exploits tracked from GitHub, ranked by stars
  • Vendor breakdown — KEV entries grouped by vendor with drill-down
  • Software search — CPE-based lookup for vulnerabilities affecting specific products
  • CWE browser — vulnerability data grouped by weakness type
  • Activity log — change detection events (new CVEs, score changes, KEV additions) with ingestion status
  • Hunt list — bug bounty relevance ranking with focus filters (PoC available, KEV, no-auth, recent)
  • CVE detail pages — full record with CVSS vector breakdown, attack surface visualization, EPSS history, PoC links, CWE/CPE data, references, and change history
  • CSV export — export filtered search results for offline analysis

Alerting

A Telegram bot sends two types of alerts. Batch reports fire after each ingestion run when new change events are detected — new critical CVEs, KEV additions, significant EPSS score spikes, or new PoC code. A daily digest runs on a configurable schedule and summarizes the highest-priority CVEs from the past 24 hours, with deduplication against previously sent alerts.

Architecture

  • Language: Python
  • Database: SQLite with WAL mode, foreign keys, and query-only mode on the dashboard connection
  • Web framework: Flask with Jinja2 templates
  • Deployment: VPS with nginx reverse proxy, systemd services for the dashboard and bot, cron-scheduled ingestion
  • Enrichment: post-ingestion enrichment pass for PoC discovery and severity reconciliation across sources
  • Change detection: tracks state changes across ingestion runs and generates typed events for alerting

Why This Matters for Vulnerability Management

This project demonstrates the core workflow of a vulnerability management program: ingest data from multiple feeds, normalize it, prioritize by real-world risk factors rather than raw severity alone, and surface actionable intelligence to the people who need it. The same pattern — collect, normalize, score, alert — scales from a solo researcher's dashboard to an enterprise vulnerability management platform.

The prioritization logic (KEV + EPSS + CVSS + PoC) mirrors what organizations like CISA recommend for patch prioritization: don't patch by CVSS alone, patch by evidence of active exploitation and exploitation probability.

View Live Dashboard All Projects