[Full-disclosure] Sony: No firewall and no patches
Bruno Cesar Moreira de Souza
bcmsouza at yahoo.com.br
Thu May 12 16:34:37 BST 2011
--- On May 11, 2011, Dobbins, Roland <rdobbins at arbor.net> escreveu:
> On May 11, 2011, at 12:52 AM, Bruno
> Cesar Moreira de Souza wrote:
> > How would you block an ACK tunnel using only a packet
> filter? (http://ntsecurity.nu/papers/acktunneling/) You don't
> need to stop the httpd service to create this kind of
> tunnel, as the packets from the attacker would just be
> ignored by the httpd service, but could be intercepted by
> the malicious code executed on the compromised server (using
> the same approach employed by network sniffers).
> See my previous response to Thor. I don't intend to
> keep this thread going forever in the face of
> incomprehension, but this focus on corner-case exfiltration
> techniques which are easily obviated by OS and service/app
> BCPs and appropriate monitoring,
I don't think it is "incomprehension". Some people just don't agree with the incorrect and generic statement that "stateful firewalls are useless to protect servers".
I can agree that in some specific cases, when the availability is the main concern for external servers, you may consider to use ACLs instead of a stateful firewall. It's an option to be more resilient against DDoS attacks. However, I can't agree that it should be a rule for every DMZ and external network in the world, because there are other options to prevent DDoS attacks (including using clustered firewalls), and also the stateful firewalls have value to restrict the action of an attacker after a server compromise.
Also, you are underestimating the skill of some attackers. My experience as a penetration tester and security incident investigator shows that it is not always so easy (even for organizations with 24x7 monitoring) to detect the action of attackers. As said before, a stateful firewall can be a strong layer of defense to restrict the damage of an attack, and to avoid backdoors and covert channels.
>to the point of
> instantiating unnecessary and harmful state in front of
> servers which makes it trivial to take them down,
I'm not convinced that it is always significantly easier to take down a firewall than the web server. In many cases, it can be also trivial to take a web server down with a DoS attack (tool example: http://ha.ckers.org/slowloris/).
> demonstrates that in general, the infosec community pretty
> much completely ignores the availability leg of the
> confidentiality-integrity-availability triad.
No, the infosec community seeks to balance these three legs. However, to decide which leg you are going to give more protection for in each environment (for example, a DMZ) and in each organization, it's better to conduct first a risk analisys.
For example, for a bank it can be worse to have an external penetration incident into critical database servers allowing modification of financial information than an unavailability incident on the Internet Banking web site.
> Which is disappointing, given that availability is in fact
> the most important leg of that triad.
This is a misleading statement. It depends on the information, environment, risks, organization etc. For many organizations, the confidentiality and integrity of the information stored in critical database servers can be much more important than the availability of the public web sites. This is why organizations with a minimum information security maturity conduct risk analisys before deciding which security mechanisms they will implement.
Bruno Cesar M. de Souza
Full-Disclosure is hosted and sponsored by Secunia.