Ah, perhaps I am the perpetrator of thread drift (or drifting consciousness
whichever you prefer). It was unintentional, and I stand corrected. Thank
you for bringing it back to the original request. When responding, I should
have done so as well. I was making an assumption about access control, and
thinking of inbound access to w2k - the destination port from the
perspective of the w2k. For outbound communication the filter, of course
could be written to only allow outbound communication using a specific
destination port and ANY source port. ANY if left to the normal
communications process would mean the use of ports above 1024, however,
perhaps a crafted packet could potentially use ANY port, including those
below.

> -----Original Message-----
> From: Scott Mulcahy [mailto:scottcmusa.net]
> Sent: Monday, November 04, 2002 12:08 PM
> To: freouwebbemsn.com; 'Scott Mulcahy'; focus-mssecurityfocus.com
> Subject: Re: [RE: Access to well-known ports on Win2K]
>
>
> I'm a bit surprised by your confusion on this topic.
>
> >> You do not need to create 1024 filters; you creat one
> which 'blocks all'
> then create one for each one you wish to allow.
> <<
>
> Normally you would use specific permits and block all else.
> If you had read
> what the original poster wanted, you would know that the
> request was for only
> admins to be able to use ANY port above 1024. This makes
> sense since client
> communication typically uses the ephemeral port range. So
> unless you intend
> to create 64,411 permit filters to allow every port above
> 1024, you'll need to
> create 1024 ports to block ports 0 - 1023.
>
> -------------------------------------
> "Roberta Bragg" <freouwebbemsn.com> wrote:
> Additional Information:
>
> TCP/IP filtering allows you to 'only' allow certain ports -
> that's blocking
> access and thats what was asked about.
>
>
> IPSEc does not provide security at the user level; its a
> machine level
> policy - works for all users of the machine; and can allow
> or block access
> according to IP address of accessing machine (to be user
> level policy, it
> would have to be able to be configured to work for a specific
> user account,
> and it cannot be)
>
>
> You do not need to create 1024 filters; you creat one which
> 'blocks all'
> then create one for each one you wish to allow.
>
> again: the policy is based on the machine; so would not
> matter if user
> logged onto domain,
>
> many routes for deployment as you mention: Group Policy;
> Local Security
> Policy; property pages of network connection; script or batch file
>
> all security is only good if its applied, as you say.
>
>
> Solaris or W2K, if a user can change the security policy
> then its not going
> to be an effective security policy.
>
> > -----Original Message-----
> > From: Scott Mulcahy [mailto:scottcmusa.net]
> > Sent: Friday, November 01, 2002 12:53 PM
> > To: focus-mssecurityfocus.com
> > Subject: RE: Access to well-known ports on Win2K
> >
> >
> > TCP/IP Filtering does not provide port level security at the
> > user level. You
> > could use an IPSec policy and deploy to all users to block
> > source ports below
> > 1024. Then have a subset of those users in a different OU
> > and assign that OU
> > a policy that permits all ports.
> >
> > There's a couple of potential problems, though. 1) You need
> > to create 1024
> > selectors for TCP (or as MS calls them filters): 1 for each
> > port. The good
> > news is you'd only have to do this once. 2) If the user is
> > able to stop the
> > IPSec Policy Agent service then the IPSec policy is no
> longer active.
> >
> > This approach has the same limitation as applying any
> > security using GPO's:
> > Problems can occur and a GPO may not get applied, if a user
> > doesn't log in
> > using AD credentials they won't get the GPO, etc.
> > Alternatively, you could
> > apply IPSec as a local security policy but management of the
> > policy just got
> > much more difficult. You would also need to make sure that
> > the user doesn't
> > have the ability to modify the IPSec policy in this case.
> >
> > Quite simply, W2K, XP, .NET have the ability to do what you
> > ask but it's not
> > as simple as Solaris. On the positive side, if you go with
> a GPO then
> > deployment is fairly simple.
> >
> > Good luck,
> > Scott
> >
> > -----Original Message-----
> > From: Roberta Bragg [mailto:freouwebbemsn.com]
> > Sent: Thursday, October 31, 2002 4:17 PM
> > To: 'Rangan, Govindaraj'; focus-mssecurityfocus.com
> > Subject: RE: Access to well-known ports on Win2K
> >
> >
> > Several well know methods for restricting port access exist
> > in WIndows 2000,
> > XP and .NET.
> >
> > Take a look at TCP/IP filtering and IPSec policies (IPSec
> > policies can be
> > written to filter port access, as well as for encrypting
> > data in flight)
> >
> > The Remote Access service can also be configured to provide
> > this type of
> > access control -
> >
> > None of these services require xtra purchases, downloads or
> > other activity -
> > they are built into the operationg system, just require
> > configuration as
> > does Solaris --
> >
> > > -----Original Message-----
> > > From: Rangan, Govindaraj [mailto:govindrti.com]
> > > Sent: Wednesday, October 30, 2002 10:59 PM
> > > To: 'focus-mssecurityfocus.com'
> > > Subject: RE: Access to well-known ports on Win2K
> > >
> > >
> > > Hi All,
> > > Greetings.
> > > Do all users on Win2K have access to the
> > > well-known ports? This
> > > question arose when I was doing some security tests in a
> > heterogeneous
> > > environment with Windows and Solaris boxes. Solaris RSHD's
> > > only security is
> > > that before allowing access, it checks the source host and
> > > source tcp port.
> > > The host should be in hosts.equiv or .rhosts and the source
> > > tcp port should
> > > be one of well known ports (0-1023). The rsh client is a
> > > setuid script and
> > > starts as root. However on Windows 2000, it is possible for
> > > any user (not
> > > necessarily an admin user) to open a "well known port" to
> > > connect to any
> > > rshd.
> > > Can we restrict access to well known ports to a
> > > certain user or
> > > group? If not, the secure way is that Solaris hosts shouldn't
> > > trust Windows
> > > hosts. Your help in resolving this is highly appreciated.
> > >
> > > Regards,
> > > Govind
> > >
> >
> >
>
>
>
>

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]