Security and software crafting for hacking minds.
 

Open the Code or Review It: Oracle CVE-2012-1675

I’m fine with Oracle, but…

This post start with the latest vulnerability CVE-2012-1675 about Oracle TNSListner.

A Man in the Middle attack is possible against the Oracle TNSListner to hijack regular users connections. The advisory says also that it is possible to make denial of services and full compromise the remote database integrity.

The attacker does not need a pair of valid credentials, it has just to reach the listener via network.

A funny way to start the week.

It all started 4 years ago

Koret, the security researcher that posted on full disclosure, first reported the issue to Oracle 4 years ago. It thought the vuln was fixed so long ago, but it doesn’t.

Oracle admitted that it was not able to provide a patch since those will ruin all regression tests for older DBMS versions that are suffering for the listener poisoning attack.

4 years ago my son yet didn’t exists, nor I was married and I was wearing Taekwondo ITF green belt.

Since 4 years your TNS Listener are exposed to a remote and critical vulnerability, that can’t be fixed in a while and the vendor was aware of it.

4 Years in the IT world is like a decade, but in the Information Security 4 years last like a century. Hopefully no TNSListners are exposed to the Internet, however consider how much database lays unprotected in flat corporate networks.

Since 4 years your TNS Listener are exposed to a remote and critical
vulnerability, that can’t be fixed in a while and the vendor was aware of it.

The quest for being certificate

You have also to consider that a lot of third-party software were build in the meantime over such vulnerable versions and were certifified to run only on certain Oracle versions.

All of those are exposed to same vulnerability and it’s not sure will certify against the new Oracle Critical Patch that will be released to fix (we all hope this time for real) the vulnerability).

A lot of third-party software were build in the meantime over such vulnerable
versions and were certifified to run only on certain Oracle versions. All of
those are exposed to same vulnerability and it’s **not sure** will certify
against the new Oracle Critical Patch that will be released to fix (we all hope
this time for real) the vulnerability).

It happens all day, some Oracle Critical patches can’t be applied because the third party vendor doesn’t certify its product to be compliant with that one.

Big vendors are just lazy and they certify compliancy with a patch usually with some months delay. But what about mid-size companies or even custom made products?

Certifications make stakeholders so comfortable about nothing bad will happen to their company.

After all, the software they paid for is certified doesn’t it?

Closed versus Open. Being certificate or just being smart?

This is not a rant about closedsource versus opensource. None in between of them carries a silver bullet for vulnerabilities or software errors.

Of course, opensource software can be theorically reviewed by a lot of smart people… but this doesn’t happen in the real world. People, even in the opensource world needs to deliver faster and smart solutions. The time to investigate a software for security issues is lacking and more important… resources lack even more.

A true fact is that when a security issue is reported in mainstream opensource project, the reaction time is misurable in days (remember: 4 years) and all is exposed and disclosed.

More important is that when a framework or a technology is released, it’s not certified to run on certain operating system version, or only if the DBMS has not installed a patch, or with certain versions of java.

A true fact is that when a security issue is reported in mainstream
_opensource_ project, the reaction time is misurable in **days** _(remember: 4
years)_ and all is exposed and disclosed.

Certifications are the most hilarious bit of all the story.

You tie your product to another product version make it immutable to changes in time. No one will ever upgrade operating system or other daemons because this would be break the certification. After each vulnerability assessment or after each new critical patch the vendor will publish, no one can upgrade because this will break the certification.

This is ridiculous. It’s like building the new facebook saying that it will run fine but you don’t have to use acer (put here the hardware manufacturer you like most) and you don’t have to access the application on Mondays.

Some rules software developers must learn from this story

  1. You write software on top on basic layers such as an operating system, a dbms or an application manager. You must people leave freedom of choice. Don’t tie your code on version x.y.z. It’s a risky behaviour and it tells the world your code is not rock solid but it inherits stability from the underlaying layers.
  2. If you really need to lock your application to a certain vendor release, please follow the vendor security patches and upgrade your certification agreement as soon as the vendor released the patch. It’s important for security teams to tell people to apply patches to mitigate a breakin exposure.
  3. Thinks in term of layers when you write your code. Most of the code can be dbms or operating system agnostic, the code it can’t be like that make it sure it will be as far as standard as possible so to be easly ported in different dbms-es or application servers or operating systems.
  4. Security is critical, for your customers and for your stakeholders. Vulnerabilities are discovered every day and software must be kept up to date to avoid security incidents. You are part of a security process, you really don’t want to stop.

An important rule for big software vendors

Independent researchers make a great work to you in sending advisories, or in finding vulnerabilities. May be they are smarter than your security team. Please read their work and close the breaches as soon as possible.

4 years is not an acceptable timeframe, moreover if the reason is that some regression test would fail.

On the other side.

The 0day exploit in this video.

Comments

Google Analytics Alternative