Liking cljdoc? Tell your friends :D

Security

Relic goes out of its way to be resistant to injection attacks with its design.

As relic's support of clojure functions allows pretty arbitrary evaluation as part of a query, e.g [sh/sh "rm" "-rf" "/"] we need to mindful of security when processing untrusted user input.

Protection offered by relic by default

relic is careful about what is allowed in the prefix function position of relic expressions (a potential vector for injection)

Relic allows:

  • function objects (not symbols!),
  • whitelisted sentinel values (that could not normally be constructed with edn/read)
  • safe keywords sentinels like :_ that cannot be used for exfiltration / eval,
  • column keywords

Out of these column keywords are the one to be worried about, although they do not provide opportunities for RCE, you should be wary of expressions like this [:extend [:my-col untrusted]] as they could be used to exfiltrate any data on the row.

I suspect in practice this is very unlikely, but see escaping if you have this problem.

I would advise in general to not allow user input to create arbitrary keywords as regardless of vectors against relic, it could be used to pollute the keyword cache and cause a DOS by increasing memory pressure / weak reference collection time.

Escaping

If you want to be absolutely sure, the :_ form can be used in expressions to ensure that the form is never interpreted and is treated as a normal value, even if a vector with a function in prefix position.

e.g [:_ [sh/sh "echo" "foo"]] would just bind the value of [sh/sh "echo" "foo"] without evaluating sh/sh.

Can you improve this documentation?Edit on GitHub

cljdoc is a website building & hosting documentation for Clojure/Script libraries

× close