As many have seen, the Log4Shell vulnerability, which was discovered over the weekend, is an extremely serious flaw and will likely impact organizations for years to come. It’s difficult to detect, used in countless software and is the perfect vehicle to get malware into your network. And unfortunately for business across the globe, there is no one cyber tool that can protect your enterprise against Log4Shell.

Instead, organizations must utilize a combination of tools and defense in-depth mindset to ensure they are prepared to detect post-compromise activity and quickly and efficiently end an attack. 

Here’s what we know thus far, and how enterprises can remain safe as this vulnerability persists.


What we know.

A member from the Alibaba Cloud Security Team reported CVE-2021-44228 (CVSS 10 out of 10) to the Log4j developers on November 24th, leading to Apache releasing a patch on December 6th. On December 9th, 2021, Apache disclosed a critical remote code execution vulnerability (CVE-2021-44228), also known as Log4Shell, which affects Apache Log4j versions 2.0-2.14.1. A PoC exploit was made public on December 10th.  It’s been suggested that the earliest evidence of exploitation they have found so far is from December 1st, meaning the vulnerability was in the wild at least 9-days before being publicly disclosed.

Log4j is a popular logging library in Java and is used in several enterprise applications, including Apache Struts, Flume, Kafka, Flink, Tomcat, Solr and VMware vCenter. Due to the prevalence of Log4j, defenders have been scrambling to identify which of their deployed applications are affected. 

Log4Shell uses the JNDI attack vector that was previously presented at BlackHat USA 2016. Exploitation of the vulnerability allows a remote attacker to execute code on the application if it logs the attacker-supplied string value with the attacker’s JNDI LDAP server lookup. To trigger the vulnerability, an attacker must include a particular string in their requests, such as user agents, which the application that uses the vulnerable Log4j library will log. The server then sends a request to the attacker’s address via JNDI. 

The attacker’s server, such as LDAP, responds by sending a path to a remote Java class file which gets injected into the vulnerable server process and executes the payload code. Additionally, administrators should be aware that threat actors can and have dumped secret data from environment variables, like AWS secret keys, to compromise the cloud.

Deutsche Telekom CERT has reported exploitation attempts originating from the Tor network. On December 10th, Imperva disclosed that they had observed upwards of 1.4 million attacks targeting CVE-2021-44228 with peaks reaching roughly 280,000 attacks per hour. Cloudflare also reported to have blocked a peak of 20,000 exploit requests per minute with around 200-400 IPs appearing to be actively scanning at any given time.


It’s okay to assume.

Assume you’ve been exposed. Assume there will be malware in your system somewhere. The broadness and evasiveness of Log4Shell illustrates this point perfectly. All managers should ask themselves if they will be able to detect whether or not attackers are using the malware to penetrate their systems.

Attempts at exploiting this vulnerability are particularly hard to detect because any string that might get logged by Log4j could trigger the vulnerability, and it could be anything from user-agent strings to email subject lines. The exploit could even be triggered down the line by, for instance, a vulnerable SIEM system that stores logs using Log4j at some later point in time.

As we know, the IT surface of modern organizations are expansive and complex. It is impossible to stop or even detect all malicious activity at the perimeter, and some will inevitably seep through the cracks. 

Defending against Log4Shell.

Defense-in-depth remains the best possible strategy for detecting Log4Shell exploitation. Mass scanning activity has already commenced with coin miners and botnets already joining the party. It is only a matter of time for ransomware affiliates to join the bandwagon. Enterprise defenders should remain vigilant and should not rely on a single detection method for detecting exploitation of this critical vulnerability. Administrators should frequently check vendors’ advisories for updates on mitigation and patch status of vulnerable products that are deployed in their enterprise.

Defense in-depth is often analyzed through the lens of the Lockheed-Martin cyber kill chain or the Mitre ATT&CK framework, which in essence both say that any successful cyberattack will have to proceed through a series of conceptually well-defined steps. In the case of Log4Shell, attackers can use it to establish initial access and install malware like Cobalt Strike and most likely the initial access will be on an internet-facing machine like an Apache HTTP web server. 

The next steps after initial access are persistence and lateral movement. Using a defense-in-depth strategy, security leaders are equipped to ask how well they are covered to detect these types of activity. For instance, it might make sense to detect persistence with host agents that report back to SIEM and to detect lateral movement with a network detection system that reports to SIEM and establishes baselines for which machines communicate with others. In this way, security teams have the ability to uncover an Apache web server that acts as a client with respect to a domain joined host behind it, which would be highly suspicious given the nature of Log4Shell. In other words, it pays to spend time and money on logging what you want instead of what you have.

With a defense-in-depth strategy and a well-thought-out logging regime, security leaders can extend security capabilities with additional software, such as user and entity behavior analytics (UEBA), which is ML-based software that can automatically generate baselines like the one in the aforementioned example. Teams can use security orchestration, automation and response (SOAR) solutions to automatically initiate an action to reduce response time and help reduce the impact of an attack, such as automatically severing the connection between the Apache web server and the domain host. The SOAR can also notify an analyst after the response action, giving the analyst the relevant information to investigate the web server and whether or not the connection should be reestablished.

Importance of detecting post-compromise activity.

Administrators must regularly check the advisories as they are being updated by the respective vendors. They can use Florian Roth’s sigma rule to detect generic exploitation attempts from web server logs. Adversaries are mostly using the user agent field for exploiting Log4Shell. However, administrators should note that the patterns of the vulnerability can be triggered by these patterns being present in any string that gets logged by Log4j. The query should be adjusted accordingly. Also, there are numerous permutations to bypass the signature, which administrators should keep in mind when using the detection.

In the current threat landscape, it is not enough for enterprise defenders to only be reactive and rely upon threat intelligence and detections. Defenders should be proactive by hunting for any suspicious activity in their environment. Even if defenders fail to detect the initial exploitation, they still have a fair chance to detect the attackers through their post-compromise activities. For starters, administrators can look out for any use of threat actors’ common tools, such as Cobalt Strike, Tor and PsExec, and if detected, proceed to determine the entire kill chain.

Finally, users can look for any anomalous server activities, like the execution of unusual processes. Administrators should note that allow listing may be required depending upon the environment.

Staying vigilant. 

The Log4Shell vulnerability is not as straightforward to exploit as it is to check for its presence, as the successful exploitation depends upon factors such as JVM version and configuration used. Nevertheless, enterprises using vulnerable applications should assume they are breached and should readily scan the application logs for any compromise artifacts. Administrators should lookout for any suspicious outbound network traffic from their vulnerable applications.