Executive brief
Vega is a declarative language for data visualization used to create interactive charts and dashboards. This vulnerability allows an attacker to execute arbitrary JavaScript code in a user's browser when they open a malicious Vega specification, potentially compromising sensitive data and enabling unauthorized actions. The attack exploits unsafe function gadgets in the global scope to bypass Vega's expression security restrictions.
Technical details
This is a cross-site scripting (XSS) vulnerability in the vlSelectionTuples function implementation within vega-selections. The root cause is insufficient validation of user-provided objects passed to vlSelectionTuples; the function unsafely calls a .map method on user-controlled objects without verifying the input is actually an array. An attacker can craft a malicious Vega JSON specification that abuses this to chain function gadgets (e.g., CanvasHandler.prototype.on via VEGA_DEBUG in the global scope) to ultimately invoke window.eval with attacker-controlled code. The vulnerability requires user interaction (opening a malicious Vega spec) and an application environment where Vega library instances or suitable function gadgets are exposed globally. The fix adds Array.isArray() validation before calling .map on untrusted input.
Affected products
- Vega vega-selections <= 5.6.1, <= 6.1.0
- Vega Vega versions using vega-selections <= 5.6.1 or <= 6.1.0
Timeline
- 2026-01-05: disclosed: Advisory published
- 2026-01-05: patched: vega-selections 6.1.2 and 5.6.3 released with fixes