XML Web Services & Security

Handy writes "Web Services (SOAP, .net, WSDL ^? , UDDI ^? ) create an even greater need for robust security. Exposed interfaces and fragmented administration coupled with a need for app-level security points to a greater need for a centralized managed security services model."

Comments

FLUFF, FLUFF, FLUFF

[newt_sd]

Not only is this article not saying a single new thing about web application security, the site at the end of the link only has 4 articles on it. This smells of advertising for a new site? Now I am not one to wear a tinfoil hat but I smell a conspiracy going on with news that isn't really news!!

Conflict of Interest?

[floppy ears]

The drive to get business advantage from XML Web Services will cause turbulent times for IT managers. To successfully navigate these new issues, managers must change their mind set from "fragmented security systems focused on using network perimeter to shield closed business systems" to "consistent managed security systems focused on managing application level security for inherently distributed business systems".

This article was written by Kerry Champion, president and Andy Yang, Senior Director of Product Management at Westbridge Technology, Inc., a provider of security and reliability infrastructure software for XML Web Services networks.
I'm not saying I disagree with their conclusion, but you always have to be suspicious when somebody comes out with an article that concludes that to be successful you have to use their product/service or something like it.

an important issue

[tps12]

I can't stress security enough. Too often we see the methodology of "write first, secure second."

No no no no. I'm sorry, that just won't cut it in today's world of scam artists. We need to be building in security on the server side from the ground up.

I am loath to resort to buzzwords, but "proactive" really describes just how I feel.

At my company we have met this challenge head-on by deploying a full server force of Mandrake Linux coupled with Apache 2. Apache 2 picks up where the original left off, with the added features of clones referring to Stormtroopers (as opposed to the original modular system). I find that our server compromises have decreased ~70% since making the switch from an IIS server farm.

I have also heard good things about BSD in regards to security and web apps. Great to see this finally getting the press it deserves.

And this is sooo different from HTTP....

[neilb78]

For a lot of companies HTTP is an exposed interface (for others it's disconnected from core data). This sounds like the same old "security on the internet" thing we've been hearing for years now.

If you can't be part of the solution there's penty of money to be made prolonging the problem. NB

SOAP = firewall bypass

[irritating environme]

I fail to see why SOAP exists except to bypass firewalls, since firewalls exist to restrict what calls/ports/protocols can be made in TCPIP. What will happen in two years will be a "firewall" system for SOAP calls, followed two years later by a new protocol to bypass that security layer, billed in an exciting acronym. Repeat ad infinitum.

You mean, like LDAP?

[GOD_ALMIGHTY]

It amazes me how much directory services are overlooked, even for this one simple use.
LDAP is made for doing centralized management. Be it user management or even configuration of services, it's built into every system and OpenLDAP is seriously robust. Just take the 10 minutes or whatever to figure out how to use LDAP and familiarize yourself with the most widely used schemas.

Using LDAP schemas is like going to create a user table in a database and having the table definition laid out for you. Also all applications should be able to follow the structure. Voila, portable services for applications.

Please, go familiarize yourself with LDAP. Not to mention SASL (RFC 2222) is meant as a system independent way of handling authentication and authorization. OpenLDAP, Cyrus IMAP and a number of other server apps handle SASL quite well, not to mention it's included in most distros.
IIRC, the Java Authentication and Authorization APIs also deal with SASL quite well.

The solutions to most of the problems that come up with 'Web Services' (a limited tool being forced on everything) have been solved by a simple trip to the IETF's RFC repository. Now you just need to use a language and environment that has libraries built for the RFC's. C or Java are your best bets, Perl comes in next, but I've found the libraries to be in various states of working, not something I'd bet my next project on.

Uh, can we say "Hailstorm?"

[Jack William Bell]

This issue is exactly why Microsoft thought they could put over Hailstorm. As a centralized model for Web Services with built-in security, user identification, preferences and certificate management Hailstorm looked like a damn good way for Microsoft to break into a new revenue space while consolidating control over the Internet.

Luckily for us Web Services weren't anywhere close to ready, at least compared to the hype for them, and Microsoft fell for their own marketing by introducing Hailstorm too soon. If they had kept it under wraps until Web Services were actually being rolled out (and running into the need for centralized security) they might have been hailed as saviors. Instead they jumped into the fray too soon and, combined with the antitrust problems, found themselves in a world of shit.

I don't know if Micrsoft has abandoned Hailstorm for good -- I do know they don't have a problem walking away from anything that doesn't pan out. But there is a chance Hailstorm, or something very similar (perhaps funded by Microsoft, but not directly owned), will return when the time is right. I expect the best model for this would be for Microsoft (and/or their competitors) to partner with the big banks and credit firms. In this case you have the businesses with the largest need for such services (and who already have significant databases) opening up their system as another revenue source. If my conjecture is valid I would expect to see announcements of such partnerships in the next six months or so.

In any case what I would like to see is an open source 'Hailstorm'. I understand there are a couple of such projects like that out there now. It would be a very Good Thing (tm) if these projects would settle on a single wire format and data model soon. Why? Because the first such system in general use is going to set the standard for everyone that follows. I would like to see both the standard itself and at least one of the implementations of that standard be open and free (as in speach).

A further extension of this concept would be to allow easy, trusted, collaboration between user identification systems. This kind of decentralization would help keep the biggies from controlling the entire dataspace. Unfortunately it may be difficult or impossible to do without compromising security.

Perhaps the best way to start is small and simple: An identification server of some kind. This service would allow you to check with with a trusted authority to make sure someone accessing your service is who they say they are. Such a server should also allow for anomynity by allowing someone to create an identity that cannot easily be traced back to the real person. Such an anomymous identify should be marked as such in some way in order to allow the service provider to decide if they want to accept it or not, but should be set up so that only the original creator of the identity can use it.

I can go on, but then I already have. Haven't I?

Jack William Bell

Re:Web Services are Flawed

[SlowMovingTarget]

While I tend to agree with your conclusion that actual Web Services (in their current formulation) are mostly a fad, I disagree with your suggestion that standardization yields weak security.

Obscure security can only secure the trivial. Lots of eyeballs checking and auditing the code, beating on it ruthlessly, does work. (That's a slight misquote Linus' Law of Debugging from "The Cathedral and the Bazaar") Witness Linux or *BSD.

Message-driven, discovery-based APIs for applications will be a good thing. For one thing, the actual APIs would fall more in line with O-O practice, where you classify objects by the messages they respond to (and not, instead, by the attributes they have). For another, you'd have a framework for service design-by-contract. But port 80 HTTP requests seem to be a lousy way to do this.

Re:Secure web services was easy years ago.

[Anonymous Coward]

Not quite that simple. Security is as much about encrypted transmission as it is making it difficult for intruders to forge their identity and pretend to be a trusted party. If a hacker gets into a trusted system, then proceeds to exploit all the 3rd parties said company works with, the vulnerability goes through the food chain.

The thing about security is, when you think you're done, you've only begun.

Centralized security management?

[Anonymous Coward]

Using insecure protocols to boot?

Geez, isn't that the same thing everyone rants on .net for?

At least with the current mishmash of crap that usually passes for any corporate network there's no "mother lode" to crack into to get the keys to the entire kingdom...