| 1 |
Oliver Schad wrote: |
| 2 |
> --------------[RFC 793 - Transmission Control Protocol]--------- |
| 3 |
[...] |
| 4 |
> --------------[RFC 792 - INTERNET CONTROL MESSAGE PROTOCOL]--------- |
| 5 |
|
| 6 |
> What was your argument? |
| 7 |
|
| 8 |
That the open and honest nature of the Internet's original design isn't |
| 9 |
always the best plan for a security policy. |
| 10 |
|
| 11 |
SMTP and HTTP proxies are open protocols as well, but it's quite |
| 12 |
inadviseable to allow global relaying for either/both service(s). |
| 13 |
|
| 14 |
I don't send anything back to any unexpected port probes because I don't |
| 15 |
want to. |
| 16 |
|
| 17 |
Sure, to some extent it is security through obscurity, but the old |
| 18 |
addage isn't entirely correct. If not for security through obscurity |
| 19 |
we'd all have our PIN numbers sharpie'd on our ATM cards. |
| 20 |
|
| 21 |
Going back to NAT - again, the old "NAT is not security!" addage is an |
| 22 |
over-generalization. One-to-many NAT *IS* inherrently secure. If some |
| 23 |
schmoe behind my NAT router decides to deploy a default install of |
| 24 |
IIS/[4-5], or SQL Server, or even an old Apache or Rsync or MySQL server |
| 25 |
- the outside world will not be able to access said service unless I |
| 26 |
give it permission (deny all, allow some is NAT's model; much |
| 27 |
consumer-level NAT equipment has an option to allow an open system, |
| 28 |
which, thankfully, I've never seen enabled per default), and I assure |
| 29 |
you, I won't until I've verified the security of the install. |
| 30 |
|
| 31 |
DROP versus REJECT isn't a be-all, end-all security mechanism. It is, |
| 32 |
however, one facet of a good security model. As long as people |
| 33 |
understand that, I don't think it's a terribly big issue. |
| 34 |
|
| 35 |
-- |
| 36 |
Stewart Honsberger - http://blackdeath.snerk.org/ |
| 37 |
To teach is to learn twice. |
| 38 |
-- Joseph Joubert |
| 39 |
|
| 40 |
-- |
| 41 |
gentoo-security@g.o mailing list |
Archives