Blog

Introducing in-browser app banners: Set guardrails for cloud apps | Learn more →

Blog
/
Identity-based attacks

Tracking identity-based attacks in the wild

To help organizations to keep track of how identity attacks are evolving, we’ve put together this helpful index of recent breaches. In particular we’re focused on tracking attacks in the public domain that demonstrate the very latest techniques being used in the wild, such as those targeting identity infrastructure itself.

Identity attacks on the rise?

Identity has been recorded as the #1 cyber attack vector since forever. You don’t have to look particularly hard to find statistics to support this. In 2023, one source reports that 4/5 breaches involved identity and compromised credentials, while another suggests that 75% of breaches are caused by mismanaged identity, access, or privileges.

Phishing, social engineering, credential stuffing, and business email compromise have morphed into a homogenous understanding of identity threats that are generally tackled through a combination of email security tooling, content access controls, and user awareness. 

Table of identity attacks - IDSA 2023. Includes phishing, social engineering, brute forcing, password spraying, credential stuffing, stolen credentials, third party attack, adversary in the middle AITM, insider attack
Breakdown of identity attacks from a 2023 IDSA whitepaper

The fact that such attacks have been reported as the top security threat for so long probably means that people pay less attention to identity threats. Ransomware grabs the headlines, and rightly so in many cases, but phishing feels like a “known known” that we have a plan for (even if the plan often fails). 

In fact, there’s a problem with messaging generally. The 2015 Verizon DBIR contains plenty of stats that still ring largely true today. For example:

  • In the 2013 DBIR, phishing was associated with over 95% of incidents attributed to state sponsored actors, and for two years running, more than two-thirds of incidents have featured phishing

  • In 60% of cases, attackers are able to compromise an organization within minutes

Remove the dates and a lot of the report still stands up. 

Bad then, worse now

But identity attacks are worse than they used to be. Yes, credential stuffing, phishing, and SIM swapping may not be the most sophisticated attacks, but they remain as effective as ever. As the saying goes, if it ain’t broke — don’t fix it. 

Recent attacks have moved toward a broader targeting of the identity infrastructure. While phishing and social engineering was once primarily a delivery mechanism for malicious payloads to be executed on endpoint, it is now used to harvest credentials and secrets for identity-based attacks against cloud apps and services. 

And because businesses have migrated to more cloud-based services and infrastructure, the compromise of an identity now has different consequences.

The data and functionality that attackers seek has moved off endpoints and internal networks and onto cloud systems and SaaS applications, which organizations are using in large numbers (tens to hundreds). The modern way of working means that applications are more often than not directly exposed to the internet — and the only thing needed to access these apps are identities. Naturally, it's much harder to stop credential stuffing attacks against 100 SaaS apps than the single centralized external VPN/webmail endpoint of yesteryear. 

It’s clear that stats alone don’t adequately capture the identity threat. So we have to look beyond the numbers to find out why. 


Using this resource

To cut through some of the noise, we’ve compiled this list of reported attacks and explored what they mean for the identity threat landscape. 

This is not intended to be an exhaustive list of all attacks involving the compromise of digital identities (the list would be endless!). Nor is it something you should read all in one go (unless you really want to, we won’t stop you). We want it to be a resource that you can refer back to, that we will continue to update as new attacks are recorded. 

In this context we define identity attacks as attacks targeting cloud identities and their associated identity management systems, protocols, applications, and infrastructure. 

The attacks recorded below are high profile examples of identity attacks that demonstrate how threat actors are leveraging the cloud identity plane to evade established cyber defenses and traverse new attack paths to achieve their goals. We’ve focused on attacks targeting identity infrastructure itself that are notable for their bypassing of traditional environments and established controls (e.g. Networkless or SaaS-to-SaaS attack paths). 

As with all publicly disclosed breaches, the level of detail and transparency we see varies. Where possible, we’ve mapped the threat actor Tactics, Techniques and Procedures to our SaaS Attack Matrix. To learn more about SaaS attack techniques read the blog.


Microsoft — January 2024

The threat group known as APT29 (also known as “The Dukes”, “Cozy Bear”, and labeled “Midnight Blizzard” by Microsoft) executed a cleverly executed password-guessing attack to compromise test cloud identities that were also lacking MFA. Attackers then leveraged this access to compromise some OAuth applications that allowed lateral movement to Microsoft’s corporate environment and the creation of other malicious OAuth applications to achieve persistence.

How did Microsoft get breached?

  • APT29 utilized password spraying / credential stuffing attacks to compromise test cloud identities that were also lacking MFA, attached to a non-production test tenant.

  • APT29 leveraged their initial access to the test tenant to identify and compromise a test OAuth application that had access to the Microsoft corporate environment by leveraging permissive Entra ID roles in the test tenant.

  • APT29 used the existing configurations to access the Microsoft corporate Entra ID tenant whereupon the app registration from the test tenant was installed as a service principal in the corporate tenant, granting the equivalent of global admin rights.

  • Using these new permissions, APT29 registered additional malicious OAuth applications in the Microsoft corporate environment, and created a new user in the Microsoft corporate tenant to grant consent to the new malicious OAuth apps, thereby achieving persistent access to the environment.

  • APT29 leveraged the elevated (maximum) privileges assigned to the ‘test’ app service principal to grant app roles to other newly created app service principals, granting them the Office 365 Exchange Online full_access_as_app role in the corporate tenant, which allows access to mailboxes.

  • APT29 leveraged these malicious OAuth applications to authenticate to Microsoft Exchange Online and target Microsoft corporate email accounts.

What was the impact of the Microsoft breach?

  • APT29 had access to Microsoft corporate email accounts, including members of the senior leadership team and employees in the cybersecurity, legal, and other functions, resulting in sensitive data leakage.

  • Microsoft has not disclosed any further impacts at this time, but it is likely that the adversary had complete, unmitigated control of the Microsoft corporate tenant for a period of time, with global administrator level access.

  • Since the initial attack there has been evidence of continued targeting, with password spraying attacks reportedly increasing tenfold, likely informed by stolen information.

What stands out in the Microsoft breach?

  • The attack was covert and targeted, with APT29 tailoring the attack to a limited number of accounts and using a low number of attempts to evade detection and avoid account blocks based on the volume of failures.

  • APT29 used residential proxy networks when interacting with the compromised tenant and, subsequently, with Exchange Online to obfuscate the source of their attack and avoid impossible travel detections. 

  • APT29 demonstrated mature and in-depth understanding of cloud infrastructure, protocols, and workflows, particularly in terms of privilege escalation and lateral movement.

  • If even Microsoft (an organization with pretty much unrivaled security resources) can’t ensure that all their accounts are protected by MFA and that there are no weak links between test/dev and prod systems, this should be a wake-up call for any company that thinks their MFA implementation is flawless. 

SaaS Attack Matrix mapping

For more information on each TTP please navigate to the GitHub entries linked in the table below. 

ID

Technique

Stage

Description

SAT1011

Credential stuffing

Initial Access

Attempt to authenticate to a SaaS account by guessing a large number of passwords

SAT1027

OAuth tokens

Execution; Persistence; Defense Evasion

Use a malicious OAuth app to create an OAuth token, using arbitrary permissions to maintain long-term programmatic access to a compromised user account.

SAT1001

Abuse existing OAuth integrations

Privilege Escalation; Lateral Movement

If an adversary compromises a SaaS account integrated with other apps, they can escalate privileges and move laterally to other apps.

Related breaches

Hewlett Packard Enterprise (HPE) — May 2023

At the time of the Microsoft breach becoming public knowledge, HPE disclosed that they had become aware of a historical incident in Dec 2023, involving unauthorized access to and exfiltration of a limited number of SharePoint files as early as May 2023. Hackers accessed and exfiltrated data from HPE mailboxes belonging to individuals in the cybersecurity, go-to-market, business segments, and other functions. No further information is available on the techniques used or impact of the breach.


Okta — October 2023

An unknown threat group compromised an Okta employee's personal Google account that was being used on a company-managed device, granting the threat actor access to a service account for Okta’s customer support system, that included session tokens for 134 customers. This was then used to hijack the legitimate Okta sessions of five customers. 

How did Okta get breached?

  • The threat actor compromised a personal Google account that the user had accessed from their Okta-managed work device by signing into their personal profile from the Chrome browser.

  • The personal account credentials are likely to have been compromised in a historical data breach and did not have MFA enabled.

  • The username and password of a service account for Okta’s customer support system had been saved into the employee’s personal Google account and was therefore compromised.

  • The threat actor was able to access the service account by logging in using the stolen credentials, which again likely did not have MFA deployed as a service account.

  • The threat actor was able to use session tokens in the HAR files to impersonate staff and hijack the legitimate Okta sessions of five customers, including 1Password, BeyondTrust, and Cloudflare.

What was the impact of the Okta breach?

  • The threat actor gained unauthorized access to files inside Okta’s customer support system associated with 134 Okta customers.

  • The threat actor was able to use these session tokens to hijack the legitimate Okta sessions of 5 (publicly disclosed) customers.

  • Okta originally claimed the breach had impacted only 1% of customers, but later found that a report run and downloaded by the threat actor contained the names and email addresses of all 18,400 Okta customer support users, as well as some Okta employee information, meaning 100% of customer support users were impacted.

  • Okta users are at higher risk of phishing and credential stuffing attacks based on the data stolen by the threat actor, increasing the importance of robust MFA implementation.

What stands out in the Okta breach?

  • This attack demonstrates the risk associated with cloud Identity Providers and the potential goldmine that they are to attackers. Much in the same way that the manufacturers of physical and virtual network appliances are continuously probed for software vulnerabilities, cloud IdPs like Okta present a huge potential opportunity, both in terms of targeting specific organizational instances as well as the Okta organization. This attack showcases the possibility of third-party supply chain attacks to target downstream organizations using IdP services. 

  • Similar to the Microsoft breach, gaps were discovered and exploited in Okta’s MFA coverage and implementation, highlighting that there are gaps in even the most mature organizations. 

  • The subsequent attack on Cloudflare (see below) and the scale of the recovery effort demonstrates the significant operational overhead in responding to and recovering from a breach of identity infrastructure, with a similar or greater scale than a traditional Active Directory compromise. While addressing the incident, Cloudflare's staff rotated all production credentials (over 5,000 unique ones), physically segmented test and staging systems, performed forensic triage on 4,893 systems, reimaged and rebooted all systems on the company's global network, including all Atlassian servers (Jira, Confluence, and Bitbucket) and machines accessed by the threat actor. All equipment in Cloudflare's Brazil data center, which was unsuccessfully targeted by the threat actor, was later returned to the manufacturers to ensure that the data center was secure.

SaaS Attack Matrix mapping

For more information on each TTP please navigate to the GitHub entries linked in the table below. 

ID

Technique

Stage

Description

SAT1011

Credential stuffing

Initial Access

Attempt to authenticate to a SaaS account by guessing a large number of passwords

SAT1028

Password Scraping

Credential Access

Collection of credentials and secrets from repositories e.g. password managers, SaaS file stores, etc.

Related breaches

Cloudflare — November 2023

The threat actor used tokens and credentials that had not been rotated to breach Cloudflare’s internal Atlassian server and access its Confluence wiki, Jira bug database, and Bitbucket source code management system. The threat actor first gained access to Cloudflare's self-hosted Atlassian server and then accessed the company's Confluence and Jira systems following a reconnaissance stage. Cloudflare says that this breach did not impact customer data or systems or the provision of services.

1Password — October 2023

1Password reported unsolicited activity in their Okta environment which was traced to a suspicious IP address. Later it was confirmed that an threat actor had accessed 1Password’s Okta environment using administrative privileges. They attempted to access the IT team member’s user dashboard, but that attempt was blocked by Okta. They also requested a report of administrative users, which was identified as suspicious and triggered an investigation. 1Password says it terminated the activity, investigated, and found no compromise of user data or other sensitive systems, either employee-facing or user-facing.

BeyondTrust - October 2023

BeyondTrust security teams detected an identity-centric attack on an in-house Okta administrator account. BeyondTrust blocked all access to the threat actor, and verified that they did not gain access to any systems. BeyondTrust has confirmed that there was no additional exposure to our internal systems or BeyondTrust’s customers.


MGM Resorts — September 2023

The threat group known as Scattered Spider socially engineered MGM help desk personnel to grant ‘super admin’ access to the Okta tenant, which was then used to steal data and deploy ransomware, resulting in significant business disruption. 

How did MGM get breached?

  • Scattered Spider researched MGM employees on LinkedIn to identify individuals likely to have privileged Okta access, specifically Super Administrator privileges. 

  • Scattered Spider contacted the IT help desk impersonating an employee with a privileged account asking for an authentication reset (password and MFA).

  • With privileged access, the compromised Super Administrator accounts were used to assign higher privileges to other accounts, circumventing MFA by removing enrolled authenticators and/or removing MFA from authentication policies.

  • Scattered Spider registered a second, attacker-controlled IdP via Org2Org using inbound federation, granting the ability to impersonate users and access applications on their behalf. By matching the username of target accounts in the second IdP to the original, the attacker was able to SSO into target applications. 

  • Through inbound federation, Scattered Spider obtained global admin rights in Azure, effectively granting full control over connected systems and granting domain admin privileges in target environments.

  • Scattered Spider deployed encryption software to around 100 ESXi servers and exfiltrated data, disrupting core business operations.

What was the impact of the MGM breach?

  • Led to a 36-hour outage of multiple MGM IT systems and affected a number of its casinos on the Las Vegas strip, including the Bellagio, Excalibur, Luxor, Mandalay Bay and New York New York.

  • Personal data compromise of an unspecified number of customers including various contact information, dates of births, genders, driver’s license numbers, social security numbers, and passport information. 

  • MGM reported that the attack would cause a $100 million hit to its third-quarter results, including $10 million in one-time cyber security consulting fees. 

What stands out in the MGM breach?

  • The MGM breach demonstrates how financially motivated organized criminal groups are specifically targeting the identity infrastructure of an organization (e.g. the chosen IdP solution) and leveraging cloud-native functionality. 

  • The MGM breach is notable for being a hybrid attack that ended in what has become a typical “actions on objective” for ransomware operators and their affiliates - the propagation of malware and encryption of core business servers. In this way attackers are leveraging the newer functionality that cloud services provide them to target non-cloud/on-premise resources. This potentially indicates that attackers see cloud applications and services as the path of least resistance to achieving their goals, exploiting more limited security team visibility and understanding of these services compared to more traditional (now well protected) targets. 

  • While attackers were focused on taking control of the cloud IdP, the initial access vector was notable for being a more traditional method (vishing) to bypass the need to acquire credentials (password and MFA token). This type of technique remains consistently effective, regardless of the technology landscape and whether MFA is correctly implemented or not.    

SaaS Attack Matrix mapping

For more information on each TTP please navigate to the GitHub entries linked in the table below. 

ID

Technique

Stage

Description

SAT1041

Inbound Federation

Persistence; Lateral Movement

Inbound federation allows users to login to a target identity provider by authenticating with a source identity provider


Retool — August 2023

Software development company Retool disclosed that the accounts of 27 of its cloud customers were compromised following a targeted SMS-based social engineering attack, which was enabled by Google Authenticator’s default synchronization of MFA tokens with the associated Google account.  

How did Retool get breached?

  • The threat actor launched a targeted SMS-based phishing campaign against Retool employees with a custom lure relating to their workplace healthcare coverage.

  • The timing coincided with a recently announced migration of logins to Okta, and the message contained a url disguised to look like their internal identity portal.

  • After logging into the fake portal – which included an MFA form – the threat actor called the employee impersonating an IT team member, deepfaking the IT employee’s real voice and using real information about the company to build trust.

  • The phished employee shared an MFA OTP token which allowed the threat actor to add their own personal device to the employee’s Okta account and enabled their own Okta MFA from that point forward.

  • Due to the Google Authenticator synchronization feature that syncs MFA codes to the cloud by default, meaning that access to a Google account immediately gave access to all MFA tokens held within that account.

  • This enabled the threat actor to take over a number of identities associated with a range of target apps and change the credentials.

What was the impact of the Retool breach?

  • A total of 27 customers were impacted, with the threat actor specifically targeting customers in the Crypto industry.

  • After taking over the accounts, the threat actor was observed gathering information and exploring the Retool apps.

  • After learning of the attack, Retool revoked all internal authenticated sessions (Okta, GSuite, etc.) for employees, locked down access to the affected accounts, notified the affected customers, and restored their accounts to their original state.

What stands out in the Retool breach?

  • Like the MGM breach, the Retool breach demonstrates how financially motivated organized criminal groups are specifically targeting the identity infrastructure of an organization (e.g. the chosen IdP solution) and leveraging cloud-native functionality. 

  • A further similarity with the MGM breach, while attackers were focused on taking control of the cloud IdP, the initial access vector was notable for being a more traditional method (SMS phishing in this case) to bypass the need to acquire credentials (password and MFA token). This type of technique remains consistently effective, regardless of the technology landscape and whether MFA is correctly implemented or not.    

  • In this case, the attacker abused inherent weaknesses in Google Authenticator, which came under fire following the breach for its default synchronization of MFA codes to the cloud when connected to an account, in order to move laterally and compromise other target apps. 

SaaS Attack Matrix mapping

For more information on each TTP please navigate to the GitHub entries linked in the table below. 

ID

Technique

Stage

Description

SAT1042

AiTM Phishing

Initial Access

Attacker-in-the-Middle (AiTM) phishing uses dedicated tooling to act as a web proxy between the victim and a legitimate login portal for an application the victim has access to, principally to make it easier to defeat MFA protection.

SAT1043

Device Enrollment

Initial Access; Persistence

Enrollment of a new MFA device in order to allow an adversary to complete MFA challenges for future authentication.


GitHub / Heroku / Travis-CI / npm — April 2022

An unknown threat actor used stolen OAuth user tokens (issued to Heroku and Travis-CI) to download data from private repositories. The threat actor then compromised an internal Heroku customer database as well as accessed and stole data from dozens of downstream organizations using Heroku and Travis-CI-maintained OAuth apps.

How did they get breached?

  • The threat actor obtained access to two third-party OAuth integrators, Heroku and Travis-CI, accessing databases and downloading stored customer GitHub integration OAuth tokens. These tokens had earlier been used by Travis-CI and Heroku OAuth applications to integrate with GitHub to deploy applications.

  • Access to the environment was gained by leveraging a compromised token for a Heroku machine account, but it is not disclosed how the threat actor achieved this. 

  • The threat actor authenticated to the GitHub API using the stolen OAuth tokens issued to Heroku and Travis CI.

  • For users who had the affected Heroku or Travis CI OAuth apps authorized in their GitHub accounts, the threat actor listed all the user's organizations.

  • The threat actor then selected targets based on the listed organizations.

  • The threat actor listed the private repositories for user accounts of interest and proceeded to clone private repositories of interest.

  • GitHub identified unauthorized access to their npm production infrastructure using a compromised AWS API key, obtained by the threat actor when they downloaded a set of private npm repositories using a stolen OAuth token from one of the two affected third-party OAuth applications.

What was the impact?

  • By stealing these OAuth tokens, the threat actor could access and download data from GitHub repositories belonging to those who authorized the compromised Heroku or Travis CI OAuth apps with their accounts. 

  • The threat actor was able to mine the downloaded private repositories for secrets that could be used to pivot to other infrastructure, stealing data from dozens of organizations. 

  • In addition to user repo’s downstream, the compromised token for a Heroku machine account obtained by threat actors also allowed unauthorized access into Heroku's internal database of customer accounts, enabling the threat actor to extract the hashed and salted passwords. 

What stands out in the Github breach?

  • Similar to the Okta breach, this attack showcases the possibility of third-party supply chain attacks to target downstream organizations using cloud SaaS services. In this case, targeting OAuth integrators as opposed to IdP providers, but with a similar goal and impact of compromising the real target organizations downstream. 

  • Applications like Github are an obvious target for attackers due to their widespread adoption. There have been numerous attacks leveraging Github as the vehicle for attacks by compromising repo’s to insert malicious code, or registering malicious copycat repo’s to dupe users into using them. 

  • Unlike the attacks abusing the functionality of Github (repo poisoning) which target the legitimate developer processes when using the app, this attack could have been prevented at the identity layer before the attacker was able to breach the Heroku/Travis-CI accounts. 

SaaS Attack Matrix mapping

For more information on each TTP please navigate to the GitHub entries linked in the table below. 

ID

Technique

Stage

Description

SAT1001

Abuse existing OAuth integrations

Privilege Escalation; Lateral Movement

If an adversary compromises a SaaS account integrated with other apps, they can escalate privileges and move laterally to other apps.

SAT1004

API keys

Persistence; Defense Evasion

An adversary that has compromised an account could then read existing API keys from the app settings, if the app allows this, or create a new API key.

SAT1006

App directory lookup

Discovery

An adversary who has gained a foothold via a SaaS app could download the list of users accessible to them in order to better target attacks against other users.


Other notable attacks

SEC X hack — January 2024

The X account for the U.S. Securities and Exchange Commission was victim to a SIM swapping attack, whereupon the attacker used the social media platform to issue a fake announcement on the approval of Bitcoin ETFs on security exchanges.

Once the threat actors controlled the number, they reset the password for the @SECGov account, and created the fake announcement. The SEC also confirmed that multi-factor authentication was not enabled on the account, as they had asked X support to disable it when they encountered problems logging into the account.


Mandiant X hack — January 2024

The X account for Mandiant was hacked by a Drainer-as-a-Service (DaaS) gang in a brute force attack. MFA was not enabled on the account. The threat actor used the social media account to share links redirecting to a phishing page to steal cryptocurrency. 

The attacker used a wallet drainer dubbed CLINKSINK. This same drainer has been used since December to steal funds and tokens from users of Solana cryptocurrency as part of a large-scale campaign involving at least 35 affiliate IDs linked to a shared DaaS.


23andMe data breach — April 2023

Genetic testing provider 23andMe confirmed that hackers downloaded the data of 6.9 million people of the existing 14 million customers after breaching around 14,000 user accounts. 

The attacker stole health reports and raw genotype data of customers affected by a credential stuffing attack that went unnoticed for five months, from April 29 to September 27. 

The credentials used by the attackers to breach the customers' accounts were stolen in other data breaches or used on previously compromised online platforms, and targeted accounts without MFA. 

Subscribe to get updates from Push
The latest news, articles, and resources, sent to your inbox weekly