Consider this:
> const a = '{"__proto__":{ "b":5}}';
'{"__proto__":{ "b":5}}'
> const b = JSON.parse(a);
{ __proto__: { b: 5 } }
> b.b;
undefined
> const c = Object.assign({}, b);
{}
> c.b
5
The problem is that JSON.parse()
retains the __proto__
property as a plain object key. By
itself, this is not a security issue. However, as soon as that object is assigned to another or
iterated on and values copied, the __proto__
property leaks and becomes the object's prototype.
Parses a given JSON-formatted text into an object where:
text
- the JSON text string.reviver
- theJSON.parse()
optionalreviver
argument.options
- optional configuration object where:protoAction
- optional string with one of:'error'
- throw aSyntaxError
when a__proto__
key is found. This is the default value.'remove'
- deletes any__proto__
keys from the result object.'ignore'
- skips all validation (same as callingJSON.parse()
directly).
Scans a given object for prototype properties where:
obj
- the object being scanned.options
- optional configuration object where:protoAction
- optional string with one of:'error'
- throw aSyntaxError
when a__proto__
key is found. This is the default value.'remove'
- deletes any__proto__
keys from the inputobj
.
Same as Bourne.parse()
except that it returns null
instead of throwing a SyntaxError
when a __proto__
key is found.