When on earth did that happen?

Can you imagine that a reputable organisation would deploy a business-critical security service without first designing and testing it, and then reviewing it to ensure that it operated as expected?  Or, would you expect an organisation to allow a security service that was not well-designed, tested and regularly reviewed to become the cornerstone of the organisation’s critical security infrastructure?

And as if those questions were not rhetorical enough, can there be any doubt about the organisation’s fate if the critical security service in question is a firewall or remote-access server?

Of course not!

So, if we accept that critical security infrastructure like firewalls and remote-access servers should be well-designed, tested before deployment and then reviewed on a regular basis, it should be implausible that similarly-critical Security Information and Event Management (SIEM), Security Operation Centre (SOC) and Incident Detection and Response (IDR) infrastructure would be treated differently!

Implausible you say?  Read on!

What happens in the hotel stays… err… nope!

Let’s have a look at just one example of how bad things can get if the SIEM/SOC/IDR doesn’t work as intended.  Back in 2019, the Marriott International hotel chain discovered that attackers had breached one of its reservation systems; that system (which originally belonged to Starwood, a hotel chain that was acquired by Marriott in 2016) was breached in 2014, and remained breached for another four years!

The main events that took place during the Starwood/Marriott breach were:

  1. July 2014 – Initial compromise of the Starwood system via a web shell in a web application called Accolade.

  2. (Unspecified method and date) The attackers gained local admin privileges. The attackers performed keystroke logging on the initial compromised server to capture passwords from 24 (privileged and non-privileged) accounts.

  3. (Multiple times in the years 2014-2018) The attackers ran Mimikatz to harvest additional credentials on multiple systems.

  4. (2015) The attackers penetrated the Cardholder Data Environment (CDE), bypassing MFA protections. The attackers were able to do this in part because certain administrator groups were excluded from Multi Factor Authentication (MFA) requirements.

  5. (2015-2018) The attackers stole data related to around 383 million customers, including over 5 million unencrypted passport numbers and over 9 million credit and debit cards.

  6. (Sept 7 2018) The attackers ran an SQL ‘select count(*)’ command on the Guest_Master_profile table which (finally) triggered an alert that was acted upon.

As we’ve noted elsewhere, IBM’s reporting shows:

  1.     The average time to identify a data breach in 2020 was 220 days, and,
  2.     Once identified, the average time to contain a data breach in 2020 was 73 days.

Of course, that 220-day average is insignificant when we see that the Starwood/Marriott data breach remained unidentified for over four years!

When on earth did that happen? From rhetoric to (expensive) reality

So, back to our first rhetorical questions:  Would you expect an organisation to allow a security service that was not well-designed, tested and regularly reviewed to become the cornerstone of the organisation’s critical security infrastructure?  Well in the case of the Starwood/Marriott breach, the plaintiffs in a multidistrict litigation (MDL) are saying that the answer is ‘no’ and their case includes specific claims regarding the effectiveness of the SIEM/SOC/IDR:

  • Only a fraction of Starwood’s firewall activity was being logged, so nobody could adequately monitor for attacks.

  • The legacy Starwood system lacked monitoring and logging of remote access, meaning that there was no record of who was remotely accessing the systems.

  • Not all database queries were being logged, so nobody could see if a hacker was accessing Starwood’s valuable data without permission.
When on earth did that happen?

No one can say what damages may be awarded should the MDL succeed but we can hark back to the 2017 Equifax breach which affected about 147 million people. Equifax’s ability to detect attackers was also raised as an issue with the Federal Trade Commission’s complaint for permanent injunction and other relief alleging: “the Defendant had minimal protections for detecting intrusions on ‘legacy’ technology systems […] which contributed to Defendant’s months-long failure to detect the attackers on its Network”.

The eventual settlement included the creation of a US$380 million fund by Equifax, payment by Equifax of US$77 million in attorney’s fees, and provision by Equifax of years worth of credit monitoring and identity protection and restoration services.

A less costly (and purple!) alternative

Call us crazy but at DotSec, we’d prefer to avoid the kind of expense and publicity that (as outlined above) accompany the failure of a SIEM/SOC/IDR service.  To achieve that goal, we initiate regular Purple Teaming (also known as Collaborative Red Teaming) exercises on both our Managed Security Information and Event Management (MSIEM) infrastructure, and against the infrastructure that is owned by our MSIEM clients.

The idea behind purple teaming is to assume a breach has already taken place. After all, a pentesting team, being limited by time/budget may not be able to identify all places where an initial compromise may take place.

Starting from this position, we build two separate groups; the Red (or attack) team and the Blue (or defence) team. The two teams cooperate (that’s very important!) to run a campaign to exercise standard Tactics, Techniques and Procedures (TTPs) in the defender’s environment with the aim of identifying which TTPs get blocked/detected by the Blue team. In concrete steps:

  1. Red and Blue teams meet, and Red shares the adversary TTPs and technical details.

  2. Blue discusses their expectations for the SIEM.

  3. Red runs the campaign.

  4. Blue follows the adversary through each TTP, noting down detection/prevention.

  5. Blue documents the results and goes away to make improvements to the SIEM/IDR.

  6. Repeat the process again.

  7. Compare the documented results and take away additional action items for improvements.

But we do penetration testing!

That’s good! Penetration testing (pen testing) can certainly help an organisation to understand how its vulnerabilities can be detected and exploited, and this is a good thing.  However for organisations that have been pen tested to death, purple teaming can be very illuminating!  For example, by starting from an ‘assume breach’ position and using TTPs which are known to both attackers and defenders, you can test your SIEM/IDR capabilities through the entire lifecycle of a breach.

Compare this to a penetration test where the main goal is to uncover vulnerabilities leading to an initial compromise, with perhaps further penetration into the target’s environment. This would be a useful exercise but it may not exercise the target organisation’s SIEM/IDR capabilities at all. And even if it does, the scenarios that are exercised will only be those that correspond to the vulnerabilities that are found and exploited by the pen tester.

When on earth did that happen?

By contrast, a purple teaming exercise assumes that a breach has taken place and is therefore focussed on exercising the SIEM/SOC/IDR capabilities of the target organisation.  And there are so many fun ways to do this!  DotSec currently has over 30 adversarial campaigns available for use, and is constantly creating new campaigns that are based on notable attacks that are making headlines.

As an example, one of our purple teaming exercises emulates an adversary that uses the following TTPs:

  • Create a backdoor to provide access back into the system at any time.

  • Search the compromised user’s files for important data.

  • Run a keylogger on the user’s machine to obtain their password.

  • Automatically learn more about the compromised company’s internal network.

  • Utilise the compromised user’s credentials to move laterally through the network.

  • Identify possible avenues to escalate privileges within the domain.

  • Encrypt files and leave a ransom note
Organisations should be confident that their SIEM/SOC/IDR systems allow effective detection and response to a campaign like this. Next time it may not be a purple team that is exercising those TTPs!

The last word!

While the Marriott MDL continues through the US court system, the UK Information Commissioner’s Office issued a fine of 18.4 million pounds (about A$33M) against Marriott (under the GDPR) for the Starwood/Marriott breach. As the ICO described, Marriott was penalised for:

  • Insufficient monitoring of privileged accounts
  • Insufficient monitoring of databases
  • Lack of controls on critical systems (particularly application whitelisting)
  • Lack of encryption on personal data

Performing regular validation of their SIEM/SOC/IDR capabilities would have definitely helped Marriott to detect and remediate at least the first three of the above four (75%!) deficiencies.

As mentioned at the beginning of this article, no one would (reasonably) expect an organisation to allow a security service that was not well-designed and/or tested and regularly reviewed to become the cornerstone of that organisation’s critical security infrastructure.

SIEM/SOC/IDR is critical security infrastructure and it needs to be regularly validated against changing IT environments to make sure it’s still going to be effective when you really need it.  Running regular Purple Teaming exercises is one of the simplest, most cost-effective (and low-publicity!) ways to achieve that goal.

PS:  It’s happened again!  This time with Neiman Marcus.

author avatar
Scroll to Top