Security & Mobility Now
Security is top-of-mind everywhere, especially right here where Juniper experts share their thoughts on the latest security breakthroughs and product advancements
OliverT

Security Trends and Futures

by Juniper Employee ‎09-06-2010 08:03 AM - edited ‎02-08-2011 05:14 PM

Over a series of blog submissions, I plan to look at the challenge of implementing security through a set of different lenses. The lens I will employ today is one of compute power – specifically the amount of compute power it takes to implement protection against a broad array of possible threats in a network device.

 

Given infinite compute capacity, one could take every flow of traffic encountered on the network and play it back (in a sandbox) against all versions of software which might end up on its receiving end. If one of the sandboxes sustains damage, we’d know that the flow is potentially dangerous (though it may involve no malicious intent and may not actually cause any damage on the particular system which is the flow’s recipient). Given that we must make do with only finite compute capacity, we will need to apply more brain and less brawn to this problem.

 

The security technologies which can prevent a flow from reaching a certain destination can be broadly categorized as follows (sequenced from coarsest and computationally cheapest to finest grained and computationally most expensive):

 

  • Network segmentation with some form of (physical or logical) admission control – the decision to place a host on a network segment (e.g. a VLAN) when it is connected to the network can be used to isolate the host from many types of potential harm.

 

  • Stateless firewall – most routers and switches can implement stateless firewall filters at line rate without breaking much of a sweat. A whole host of undesirable behaviors can be headed off with stateless firewall rules which prevent traffic which absolutely shouldn’t be on your network from being there.

 

  • Stateful firewall – stateful firewall can protect against more complicated protocol anomalies that have historically been exploited to compromise systems. In the past, TCP has been a favorite target in the protocol stack and stateful firewalls have grown to meet this challenge.

 

  • L5-L7 protocol anomaly detectors – stateful firewalls detect anomalies in layers 3 (IP) and 4 (UDP, TCP, etc.), while products such as Intrusion Prevention Systems (IPS) include (along with other capabilities) higher level protocol decoders (e.g. HTTP) which can detect anomalies and malformations of those protocols. Other examples in this class include XML firewalls, which can verify that XML documents are well-formed, are compliant with a specified XML schema and don’t do silly things like nest a thousand levels deep.

 

  • Signature-based pattern matching – examples include the signature-based portion of IPS systems which can detect a variety of attacks and XML firewalls which look for SQL injection exploits. These technologies become computationally expensive enough that much energy is expended trying to optimize them (the more context that is available, the more they can be optimized).

 

  • Behavioral analysis – establishing some sense of what “normal” looks like and looks for significant variations from this baseline. With fine granularity, it can be quite expensive to implement.

 

  • Test Execution – the final step involves executing the traffic pattern in a sandbox. This is more viable for detecting attacks in things like JavaScript, where the target execution environment is somewhat constrained.

The best network security architectures make use of all of these tools – each in its appropriate place. And the layers which these technologies represent can be highly symbiotic: if a test execution finds an offending piece of JavaScript (a computationally expensive operation), it can compose and install an IPS signature further up the stack (and computationally cheaper) to catch subsequent instances of the attack.

 

It’s not worth wondering which security technology you need – each is a tool in a tool chest and limiting your selection of tools puts you in the awkward situation not unlike trying to assemble a grill for your labor day picnic with only a hammer in hand.

Comments
by Sanjay Gupta(anon) on ‎09-28-2010 04:00 PM

Oliver,

 

Nice overview .. I think it would be valuable to add some quantitative perspective on the same to get a deeper understanding. By that I mean lets quantify what kind of compute power jump do we need as we go from one kind of security technology to other? Here is the outlook from $$ perspective (very rough):

 

stateless vlan based/ACL based security at ~$6K/10GE on EX (L2), ~12K/10GE (L3)

stateful FW (L4) ~ $50k/10GE

IDP/IPS  ~120k/10GE

 

So we see a jump of almost 5x-6x going from stateless to stateful fw and then 2x-3x jump from stateful fw to deeper inspection (quite fuzzy here since there are lot of variation to it)

 

Another dimension is from a deployment perspective, though as you said each of the above tools have their use, it would be interesting to understand how often or how much do we see of each and where? 

by thailand ccna voice training (anon) on ‎08-03-2012 04:12 AM

Firell is a must for a network. 

<a href="http://internetworksolutions.net/cisco-training/cisco-ccna/ccna-voice-training/
">thailand ccna voice training
</a>

Post a Comment
Be sure to enter a unique name. You can't reuse a name that's already in use.
Be sure to enter a unique email address. You can't reuse an email address that's already in use.
Type the characters you see in the picture above.Type the words you hear.
About Security & Mobility Now

Discussing a wide range of topics impacting enterprises and
data center security.

Subscribe RSS Icon

Our Bloggers

Kyle Adams
Senior Software Engineer

Profile | Subscribe

Ritesh Agrawal
Director
Software Engineering

Profile | Subscribe

Erin K. Banks
Senior Technical Marketing Manager

Profile | Subscribe

Ajay Bharadwaj
Product Manager

Profile | Subscribe

Michael Callahan
Vice President
Product Marketing

Profile | Subscribe

Scott Emo
Director
Product Marketing

Profile | Subscribe

Mora Gozani
Senior Manager
Product Marketing

Profile | Subscribe

Ashur Kanoon
Sr. Manager
Technical Marketing

Profile | Subscribe

Seema Kathuria
Manager
Product Marketing

Profile | Subscribe

Kevin Kennedy
Senior Director
Product Management

Profile | Subscribe

Dave Killion
Software Engineer

Profile | Subscribe

Rebecca Lawson
Senior Director
Product Marketing

Profile | Subscribe

Rajoo Nagar
Senior Manager
Product Marketing

Profile | Subscribe

Erin O'Malley
Manager
Product Marketing

Profile | Subscribe

Galina Pildush
Strategy & Planning
Architect

Profile | Subscribe

Edward Roberts
Director
Product Marketing

Profile | Subscribe

Bill Shelton
Director Field Sales

Profile | Subscribe

Ashutosh Thakur
Product Line Manager

Profile | Subscribe

Troy Vennon
Software Engineer

Profile | Subscribe

Brad Woodberg
Product Manager

Profile | Subscribe

Labels
Copyright© 1999-2013 Juniper Networks, Inc. All rights reserved.