CVE-2026-33744

BentoML has Dockerfile Command Injection via system_packages in bentofile.yaml

Summary

The docker.system_packages field in bentofile.yaml accepts arbitrary strings that are interpolated directly into Dockerfile RUN commands without sanitization. Since system_packages is semantically a list of OS package names (data), users do not expect values to be interpreted as shell commands. A malicious bentofile.yaml achieves arbitrary command execution during bentoml containerize / docker build.

Affected Component

• src/_bentoml_sdk/images.py:85-89 — .format(packages=" ".join(packages)) into shell command
• src/bentoml/_internal/container/frontend/dockerfile/templates/base_debian.j2:13 — {{ __options__system_packages | join(' ') }}
• src/bentoml/_internal/bento/build_config.py:174 — No validation on system_packages
• All distro install commands in src/bentoml/_internal/container/frontend/dockerfile/__init__.py

Affected Versions

All versions supporting docker.system_packages in bentofile.yaml, confirmed on 1.4.36.

Steps to Reproduce

• Create a project directory with:

service.py:

import bentoml
@bentoml.service
class MyService:
    @bentoml.api
    def predict(self) -> str:
        return "hello"

bentofile.yaml:

service: "service:MyService"
docker:
  system_packages:
"curl && id > /tmp/bentoml-pwned #"

bentoml build

• Examine the generated Dockerfile at ~/bentoml/bentos/my_service//env/docker/Dockerfile. Line 41 will contain:

RUN apt-get install -q -y -o Dpkg::Options::=--force-confdef curl && id > /tmp/bentoml-pwned #

• Running bentoml containerize my_service: will execute id > /tmp/bentoml-pwned as root during the Docker build.

Root Cause

The system_packages field values are treated as package names (data) by the user but are string-formatted directly into shell commands in the Dockerfile:

# images.py:85-89
self.commands.append(
    CONTAINER_METADATA[self.distro]["install_command"].format(
        packages=" ".join(packages)  # No escaping
    )
)

Where install_command is "apt-get install -q -y -o Dpkg::Options::=--force-confdef {packages}".

A bash_quote filter (wrapping shlex.quote) exists in the codebase and is registered in both Jinja2 environments, but it is only applied to environment variable values, never to system_packages.

Impact

• Malicious repositories: An attacker publishes an ML project with a crafted bentofile.yaml. Anyone who clones and builds it gets arbitrary code execution during docker build.
• CI/CD compromise: Automated pipelines running bentoml containerize on PRs that modify bentofile.yaml are vulnerable.
• BentoCloud: If BentoCloud builds images from user-supplied bentofile.yaml, this could achieve RCE on cloud infrastructure.
• Supply chain: Shared bentos or model repos in the BentoML ecosystem can contain malicious configs.

Suggested Fix

Option 1: Input validation (recommended)

Add a regex validator to system_packages in build_config.py:

import re
VALID_PACKAGE_NAME = re.compile(r'^[a-zA-Z0-9][a-zA-Z0-9.+\-_:]*$')
def _validate_system_packages(instance, attribute, value):
    if value is None:
        return
    for pkg in value:
        if not VALID_PACKAGE_NAME.match(pkg):
            raise BentoMLException(
                f"Invalid system package name: {pkg!r}. "
                "Package names may only contain alphanumeric characters, "
                "dots, plus signs, hyphens, underscores, and colons."
            )
system_packages: t.Optional[t.List[str]] = attr.field(
    default=None, validator=_validate_system_packages
)

Option 2: Output escaping

Apply shlex.quote() to each package name before interpolation in images.py:system_packages() and apply the bash_quote Jinja2 filter in base_debian.j2.