simple-squiggle/node_modules/mathjs/docs/expressions/security.md

90 lines
4.1 KiB
Markdown
Raw Normal View History

# Security
Executing arbitrary expressions like enabled by the expression parser of
mathjs involves a risk in general. When you're using mathjs to let users
execute arbitrary expressions, it's good to take a moment to think about
possible security and stability implications, especially when running
the code server side.
## Security risks
A user could try to inject malicious JavaScript code via the expression
parser. The expression parser of mathjs offers a sandboxed environment
to execute expressions which should make this impossible. It's possible
though that there are unknown security vulnerabilities, so it's important
to be careful, especially when allowing server side execution of
arbitrary expressions.
The expression parser of mathjs parses the input in a controlled
way into an expression tree or abstract syntax tree (AST).
In a "compile" step, it does as much as possible preprocessing on the
static parts of the expression, and creates a fast performing function
which can be used to evaluate the expression repeatedly using a
dynamically passed scope.
The parser actively prevents access to JavaScripts internal `eval` and
`new Function` which are the main cause of security attacks. Mathjs
versions 4 and newer does not use JavaScript's `eval` under the hood.
Version 3 and older did use `eval` for the compile step. This is not
directly a security issue but results in a larger possible attack surface.
When running a node.js server, it's good to be aware of the different
types of security risks. The risk whe running inside a browser may be
limited though it's good to be aware of [Cross side scripting (XSS)](https://www.wikiwand.com/en/Cross-site_scripting) vulnerabilities. A nice overview of
security risks of a node.js servers is listed in an article [Node.js security checklist](https://blog.risingstack.com/node-js-security-checklist/) by Gergely Nemeth.
### Less vulnerable expression parser
There is a small number of functions which yield the biggest security
risk in the expression parser:
- `import` and `createUnit` which alter the built-in functionality and
allow overriding existing functions and units.
- `evaluate`, `parse`, `simplify`, and `derivative` which parse arbitrary
input into a manipulable expression tree.
To make the expression parser less vulnerable whilst still supporting
most functionality, these functions can be disabled:
```js
import { create, all } from 'mathjs'
const math = create(all)
const limitedEvaluate = math.evaluate
math.import({
'import': function () { throw new Error('Function import is disabled') },
'createUnit': function () { throw new Error('Function createUnit is disabled') },
'evaluate': function () { throw new Error('Function evaluate is disabled') },
'parse': function () { throw new Error('Function parse is disabled') },
'simplify': function () { throw new Error('Function simplify is disabled') },
'derivative': function () { throw new Error('Function derivative is disabled') }
}, { override: true })
console.log(limitedEvaluate('sqrt(16)')) // Ok, 4
console.log(limitedEvaluate('parse("2+3")')) // Error: Function parse is disabled
```
### Found a security vulnerability? Please report in private!
You found a security vulnerability? Awesome! We hope you don't have bad
intentions and want to help fix the issue. Please report the
vulnerability in a private way by contacting one of the maintainers
via mail or an other private channel. That way we can work together
on a fix before sharing the issue with everybody including the bad guys.
## Stability risks
A user could accidentally or on purpose execute a
heavy expression like creating a huge matrix. That can let the
JavaScript engine run out of memory or freeze it when the CPU goes
to 100% for a long time.
To protect against this sort of issue, one can run the expression parser
in a separate Web Worker or child_process, so it can't affect the
main process. The workers can be killed when it runs for too
long or consumes too much memory. A useful library in this regard
is [workerpool](https://github.com/josdejong/workerpool), which makes
it easy to manage a pool of workers in both browser and node.js.