Originally written by Taha Karim

CVE-2019–1367 background and in-the-wild exploitations

There are some important aspects to know about CVE-2019–1367 before diving into the technical analysis including the intel around it and the series of events following the in-the-wild exploitations and Microsoft patches.

First of all, here is the bug class of this bug based on the Google P0 report:

JScript variable (represented as VAR structure) isn’t properly tracked by garbage collector

Is it important to understand the bug class. For example Google P0 will work on what they call a Variant Analysis, to discover additional vulnerabilities from the same bug class. Example CVE-2019–1429 is a result of Variant analysis of CVE-2019–1367.

In fact, the first in the wild exploitation from this bug class was seen in December 2018, exploiting CVE-2018–8653, discovered by Google TAG Team. This bug was then documented by Mcafee here and by Tetrane here.

Google TAG Team discovered CVE-2019–1367 exploited in the wild by a threat actor. No details were given at the time of Microsoft advisory.

But in a recent blog Google TAG Team discussed that North Korea or individuals who worked on North Korea-related issues were the main targets of this in-the-wild exploitations but no more elements were given regarding the threat actors behind these attacks at the time of the reporting.

Who are these threat actors?

Based on OSINT, and the data following this discovery, it seems that there are two main threat actors known to date caught exploiting CVE-2019–1367:

DarkHotel APT: A suspected Korean Peninsula APT actor, considered to be a skill-full, active, long-run (+10 years of existence) resourceful (state sponsored?) APT actor.

Magnitude Exploit KIT: An opportunistic Malvertiser, mostly targeting south Korea. Magnitude EK has been there since 2013, and known to drop very known ransomware families including: Locky, Cerber, Magniber, CryptoWall, GranCrab.. Known to rapidly integrate CVE’s into their exploitation chains. Magnitude has been active for many years, below is an tweet from 2013 showing some of their oldest CVE integrations:

So what’s CVE-2019–1367 impact?

CVE-2019–1367 enables Remote Code Execution (RCE) in the context of Internet explorer in all version from 8, 9, 10 and 11 due to a memory corruption in jscript.dll.

This means victims will ultimately be infected with a malware just by browsing to a web page aka 1-click exploit : Victims need to click at least one time into a link to get infected (different from 0-click exploits that requires no user interaction at all).

A scenario of 1-click exploit attack would be of watering hole type of attack, this is usually a technique of nation state APTs.

Note: most recent iOS exploits exploited in the wild were found in watering hole attacks, targeting certain populations.

We also see 1-click exploit used in malvertising (malicious ads) mostly by cyber crime/FIN actors. In the scenario where 1-click exploits are integrated into a chain of redirects like we see everyday in malicious advertising attack, this 1-click exploits could have a 0-click effect, since no user interaction will be required before getting infected.

Even though 1-click exploits are less valuable than 0-click exploits (and less pricey), they can have a similar devastating effect if integrated at the right place.

Microsoft released a patch and encouraged users to disable jscript.dll (a Legacy dll replaced by jscript9.dll) that can still be called with IE-8 compatibility mode enabled. This is typically enabled via the following tag:

<meta http-equiv=”X-UA-Compatible” content=”IE=8”></meta>

To give more context regarding CVE-2019–1367 we draw a timeline of events, that we collected (based on OSINT)

Timeline of the events

24th September 2019: @_clem1 from Google TAG Team acknowledged by Microsoft. Darkhotel APT first time cited as linked to the in the wild exploitation of this bug:

24th September 2019: Kaspersky confirmed that DarkHotel APT was exploiting this vulnerability:

Based on the above tweet, Google Project Zero maintains a google doc referencing 0-day in-the-wild exploitation where they officially attributed CVE-2019–1367 to DarkHotel APT.

20 January 2020: Samples from DarkHotel APT exploits were uploaded to VirusTotal and flagged as CVE-2019–1367 by most of the security vendors:

6 February 2020: Google Project Zero did a Variant Analysis of CVE-2019–1367 which resulted in CVE-2019–1429 which a variant of the same bug class:

In their presentation they gave indication about the root cause of CVE-2019–1367 bug, which is as following:

Function arguments not being tracked by garbage collector during Array.sort callback, in jscript.dll.

So now we have a good lead on what type of object and function we need to look for this exploit analysis.

11 February 2020: Quihoo 360 published details about DarkHotel APT switching to another 0-day exploit, the attack was called “Double Star” as it involves this time a 0-day in Internet Explorer (CVE-2020–0674) and a 0-day in FireFox (CVE-2019–17026). Same day as Microsoft published the advisory. No more additional details were given about the vulnerability.

Based on Google P0 tweet below, it seems that CVE-2020–0674 is patching a misfix of the earlier CVE-2019–1367 suggest that CVE-2020–0674 is a variant from the same bug class:

29th February 2020: Kang Yang(@dnpushme) of Qihoo 360 ATA found that the original proof of concept they sent to Microsoft for CVE-2020–0674 was uploaded to VirusTotal. ie, the proof of concept is now available to the security community!

After analysis of this “leak” we confirmed that CVE-2020–0674 is same bug class and is a variant of CVE-2019–1367 :

3rd March 2020: Magnitude Exploit Kit actors caught exploiting the vulnerability in the wild targeting south Korea, based on a report of Ahnlab (a south Korean Security Vendor)
4th March 2020: While analyzing recent Exploit KIT attacks from different OSINT sources, we stumbled upon a recent Magnitude Exploit KIT attack targeting what seems to be Korean Users. The pcaps of a successful attack were generously provided by malware-traffic-analysis!

With bluegas[.]website hosting exploit code and pophot[.]website hosting ransomware:

After our analysis of the pcaps files, the attack turns out to be exploiting CVE-2019–1367! So we have now a proper source to analyze this exploit.

26th march 2020: Google TAG Team confirmed in their blog report that CVE-2018–8653, CVE-2019–1367 and CVE-2020–0674 are vulnerabilities inside jscript and are abusing the Enumerator object. They also give some indications about threat actor exploiting these vulnerabilities in the wild. Google stated that attack was targeting North Korea or individuals who worked on North Korea-related issues. This is inline with Kaspersky reporting above, and in line with Magnitude Exploit Kit Targeting Korean users.
2nd April 2020: Japan CERT confirmed attacks on japan infrastructures involving “Double Star” exploits, attack attributed again to DarkHotel APT.

With last.tax-lab[.]net domain hosting exploit code and backdoor

14th April 2020: Another related scripting engine vulnerability patched by Microsoft, caught exploited in the wild by Google. Microsoft assigned CVE-2020–0968, We do not know if this bug is from the same bug class as CVE-2019–1367 and CVE-2020–0674 , as no details were provided by Google or by Microsoft:

8th May 2020: @maxpl0it from F-secure released a plain CVE-2020–0674 proof of concept exploit on Github for windows 7 x64 :

15th May 2020: First write up of CVE-2020–0674 by F-secure labs, explaining the vulnerability and ways to exploit in x64 systems. This confirms our assumptions of CVE-2020–0674 being a variant of CVE-2019–1367 (or a missfix depends on which angle we see it).

5 June 2020: @_clem1 from Google TAG Team gave a presentation on 0-day exploitation in the wild and shared and for the first time (AFAIK) important differences between CVE-2019–1367 and CVE-2020–0674 bugs. This confirmed directly some of our assumptions after analyzing the POC of CVE-2020–0674 that was uploaded to VT see (above timeline in 29th February 2020), and clearly helped us differentiating between the two bugs.

Elements we have so far:

The effort spent in the intel was worth it. At this point we are armed with the following:

We have pcap files from a successful CVE-2019–1367 exploitation.
Proof of concept code of CVE-2020–9674.
We know the bug class, and the vulnerable code path for both CVE-2020–0674 and CVE-2019–1367.

CVE-2019–1367 analysis

For this analysis, we will start writing a POC for the CVE-2019–1367 vulnerability in part1.

In part2 we will discuss how this vulnerability was exploited by Magnitude Exploit Kit and DarkHotel APT and shed some lights on the differences.

Finally part3 is the shellcode analysis resulting from a successful exploitation of this bug in the wild.

Below is the table of content broken-down into different parts for readability

Table of Contents:

CVE-2019–1367 background and in-wild exploitations (Prelude)
Vulnerability Analysis + Proof of Concept (part 1)
Exploitation of CVE-2019–1367 (part 2)
Magnitude Exploit Kit Shell code Analysis (part 3)

Tools used of the analysis:

IDA Pro v7.5
WinDBG: TTD feature, javascript engine
radare2 : r2pipe
Python3
Microsoft Debug symbols.

Happy reading !

Internet Explorer CVE-2019–1367 In the wild…