[erlang-questions] Erlang offensive paper

[Richard A. O'Keefe]

On 2/06/16 10:26 AM, Lee Sylvester wrote:
> In fairness; should we ever rely on the underlying virtual machine to 
> be secure for any platform?

Can we?  Maybe, if there is a formal security model and the VM has been
verified against that model.  Should we?  No.  As a long-time reader of
comp.risks, I keep on being amazed/amused/horrified/driven to despair
by the weird and wonderful ways people find to break things, that I could
never begin to imagine, let alone prevent.
> If you were coding a *ahem* NodeJS app, would you rely on its security?
If I were coding a node.js app, I would need my university's Employee
Assistance Program's 3-free-sessions-with-a-counsellor to deal with 
depression.

> Personally, I implement security for the messaging, be it HTTP or 
> sockets etc., and implement safeguards around that VM through other 
> technologies.  In fact, I even proxy my HTTP / sockets.
You know the joke about how to secure a computer.
It takes a man and a dog.
The man is there to feed the dog.
The dog is there to bite the man if he tries to turn the computer on.

You can do everything you said and your system will still leak information
like crazy, see TEMPEST.  And of course you have to trust your OS,
hypervisor, and HAL, which I don't.

The practical question is how MUCH vulnerability Erlang adds compared
with alternatives, and how much it costs to mitigate those vulnerabilities
compared with alternatives.  The paper that triggered this thread (or at
least the slides) didn't present any numbers.  I haven't any myself.