Abstract
Modern software development is heavily reliant on open-source software (OSS) dependencies, which significantly expands an organization's attack surface. Managing these vulnerabilities across thousands of repositories is a major challenge for security analysts. This paper introduces VulnEx (Vulnerability Explorer), a visual analytics tool designed to provide an organization-wide overview of security risks. By shifting focus from individual applications to the entire portfolio, VulnEx enables analysts to identify widespread vulnerable libraries and prioritize remediation efforts effectively.
Introduction
The reliance on open-source libraries speeds up development but introduces systemic risks. A single vulnerability in a popular library can compromise hundreds of applications within a large organization. Current security workflows often rely on text-heavy reports focused on individual projects, leaving a gap in organizational-level visibility. Security analysts need to answer critical questions: Which projects are most at risk? Which library causes the most trouble across the portfolio? Is a specific high-profile CVE present in any system? VulnEx addresses these needs through a unified visual interface.
The VulnEx Visual Analytics Tool
VulnEx utilizes a "tree+table" visualization approach to bridge the gap between high-level overviews and technical details. Developed through a user-centered design process with security experts at SAP, the tool offers three complementary perspectives:
- Repository-centered: Focuses on applications and their specific risks.
- Library-centered: Focuses on dependencies and their prevalence across the organization.
- Bug-centered: Focuses on specific CVEs to track their footprint.
Visual Encodings and Metadata
The tool employs heatmaps to represent the distribution of CVSS (Common Vulnerability Scoring System) scores, allowing for quick identification of "hotspots." A CVE Matrix shows the co-occurrence of specific bugs across different entities.
Beyond security scores, VulnEx integrates "health" metrics from GitHub (stars, watchers, issues) and LGTM (code quality grades). This metadata helps analysts assess whether a library is well-maintained or a "dead" project that should be replaced, reflecting a shift in thinking where library health is a component of security.
Evaluation and Key Insights
The authors evaluated VulnEx using a dataset of 295 public GitHub repositories from the Eclipse Foundation, focusing on Apache Maven projects written in Java. Vulnerabilities were identified using Eclipse Steady.
The Power of Aggregation
The evaluation demonstrated that vulnerability management is an optimization problem. For instance, identifying that a single library like activemq-all affected 20 different repositories showed that one patch could significantly lower the organization's overall risk profile.
Preference for Categorical Data
Security experts preferred categorical heatmaps (Low, Medium, High, Critical) over precise numerical CVSS values. For high-level auditing, the qualitative severity ("how bad is it?") proved more actionable than minor numerical differences (e.g., 9.2 vs 9.3).
Legacy Risk Identification
In the Eclipse dataset, VulnEx quickly identified that the low-marmoset repository was exposed to multiple critical and high-severity vulnerabilities. It also helped find "ancient" bugs, such as a medium-severity vulnerability from 2009 (CVE-2009-2625) still present in 27 repositories.
Conclusion and Future Work
VulnEx provides a much-needed "big picture" for security auditing and governance. While effective for hundreds of repositories, future work will address scalability for organizations with over 1,000 repositories. The authors also plan to incorporate temporal analysis to track risk exposure over time and expand language support beyond Java and Python to ecosystems like npm. Collaborative features, such as annotating false positives, are also under consideration to enhance team workflows.
Summary updated