Huge Increase in Scanning for CVE-2017-9841 With Large Variability in Scanning Infrastructure

Introduction

Welcome to the June 2024 installment of the Sensor Intelligence Series, our monthly summary of vulnerability intelligence based on distributed passive sensor data.

We observed a massive increase in scanning for CVE-2017-9841, continued increases in scanning for CVE-2023-1389, and scanning for a newly discovered PHP vulnerability – CVE-2024-4577.

CVE-2023-1389, an RCE vulnerability in TP-Link Archer AX21 consumer routers, continues to rise (up 33% from last month). This has been our top scanned vulnerability since March and would be on top again this month if CVE-2017-9841 figure had not suddenly leapt to the top spot.

CVE-2024-4577 is a newly discovered vulnerability in PHP when running on Windows using Apache and PHP-CGI.

CVE-2017-9841 is a very old vulnerability in PHPUnit, which we’ll look at in detail next.

CVE-2017-9841 Scanning Source Analysis

The massive increase in scanning this month for CVE-2017-9841 strikes us as very odd. Why would scanning for this 7-year-old vulnerability suddenly spike? It seems unlikely that there are a lot of vulnerable, unpatched instances of PHPUnit at this point, at least ones that haven’t already been compromised. It could be that someone has noticed a new product or piece of software includes a vulnerable version of this code, but we have found no evidence of this.

It is also possible, indeed even likely, that unpatched versions of PHP with this vulnerability can be compromised many times. The specifics of this vulnerability are quite simple – vulnerable versions will execute data provided to them in a POST requests so long as the data starts with ‘<?php’. Nothing about this would prevent multiple compromises, and this may, depending on the attacker’s agenda, be more than enough reason to target this vulnerability.

We dug into the sources of these scans and looked at what else they are targeting and found some interesting information.

First off all, scanning for this vulnerability has been present in our dataset from the very beginning of this project, all the way back in 2020. The following table shows the number of scans detected across our entire data set, by year.
 

Scanning peaked in 2021 and decreased in 2022, but in just the first six months of 2024, this situation changed, with 100,607 events observed. Breaking this out by month shows very clearly the massive increase, and adding additional fields reveals some interesting patterns.

Note the large increase in the number of unique source IPs and source ASNs. Between May and June, 38 different source ASNs dropped from the scanning activity, and 179 were added. This is unusual. While scanners will abandon infrastructure as takedowns happen, or access is revoked, they typically do not make such massive changes without need. We suspect that this is may be a case where the directors of this scanning activity have decided that they require far more resources and infrastructure than they have had before, and have further decided to diversify their sources. Why they would do this remains a mystery, but perhaps they are building in resiliency against actions against them.

It is interesting to note that most of the scanning infrastructure is in just a handful of countries – specifically the US (35 ASNs), China (20 ASNs), Russia (18 ASNs), Hong Kong (17 ASNs), Germany (14 ASNs), Vietnam (12 ASNs) and Singapore (11 ASNs). Of the remaining countries, 23 had ASN counts greater than one, and the rest of the countries, 19 in all, had just one ASN involved.

Nevertheless, the breadth of source ASNs is somewhat astonishing – many of the scanners we track have just a handful of ASNs in use, but this group has a large geographic distribution, with infrastructure in ASNs in Kazakhstan, Moldova, the Seychelles, and Cyprus, among others.

It is difficult to ascertain if this is all the activity of one actor or a group of unrelated ones. We surmise that given the consistency of the targeting (hitting only about half of the sensors we have deployed from a country-by-country basis), as well as the unusual consistency of only one set of headers that this is likely just one actor, but we can’t be 100% certain.

Looking at what this set of scanners is looking for, we find an emphasis on variations of the PHPUnit vulnerability. However, these scanners are also doing some other types of scans, specifically “credential finding” type scans, looking for unsecured “.env” files, git configuration directories, and exposed log files, although at drastically lower levels.
 

We don’t currently understand the reasons behind the intensity of the scanning for this vulnerability, nor the dramatic increase in scanner infrastructure but clearly something is going on. We’ll keep digging into this and see what we can find for next month.

June Vulnerabilities by the Numbers

Figure 1 shows June attack traffic for the top ten CVEs that we track. Note the continued presence of CVE-2023-1389, but also the enormous amount of scanning for CVE-2017-9841, continuing a rise that’s been increasing, albeit slowly, since March 2024.

We also note the appearance of a very recent PHP vulnerability, CVE-2024-4577. This is a vulnerability that allows both disclosure of source code and remote code execution on vulnerable versions of PHP running on Windows operating systems using Apache and PHP-CGI. This was announced by Orange Tsai on Friday, June 7^th, 2024,¹ and we first saw it show up in our scanning logs on the very same day, at 20:10:35 GMT.

Other changes to the top ten include the decline in scanning for CVE-2019-9082, as well as the appearance of CVE-2022-47945 in our top ten. This vulnerability, a ThinkPHP Framework local file inclusion bug, has been in our logs for a long time, but has experienced significant growth in the last month.

Targeting Trends

Figure 2 is a bump plot showing the change in traffic volume and position over the last twelve months. This shows very clearly the massive increase in scanning for CVE-2017-9841. Also notable is the continued increase in scanning for CVE-2023-1389, with its total volume (seen in the width of the colored area) indicating that more scanning for this occurred last month than at any time in the previous eleven months.

Long Term Trends

Figure 3 shows traffic for the top 19 CVEs by all-time traffic, followed by a monthly average of the remaining CVEs. This once again shows the dramatic increase in CVE-2017-9841, as well as the movement of CVE-2023-1389, along with modest increases in CVE-2019-9082, and CVE-2022-47945.

Conclusions

Despite the oddities of this month, we continue to see an intense focus on the compromise of IoT devices, with the goal of assembling massive global botnets. As we have said many times before, IoT is a favorite target of attackers, as they are frequently forgotten, unpatched, and easily accessible. They provide deniable infrastructure for further activities, and frequently (especially in the case of consumer routers) on residential networks, a boon when operating against companies or government organizations which must provide access to their customers.

We’ll also suggest that the presence of CVE-2024-4577 reminds us that a serious vulnerability in any software package can be found at any time, and monitoring, patching, and mitigating such vulnerabilities is a continuous process.

We hope you enjoyed this deep dive into a single instance of scanning activity, and we’ll see you next month!