Stealing more than half a million euro in just a week - it sounds like a Hollywood heist movie. But the organizers of the Luuuk banking fraud pulled it off with a Man-in-the-Browser (MITB) campaign against a specific European bank. The stolen money was then automatically transferred to preset mule accounts. When GReAT discovered Luuuk's control panel it immediately got in touch with the bank and launched an investigation.
On January 20th 2014 Kaspersky Lab detected a suspicious server containing several log files including events from bots reporting to a command and control web panel. The information sent seemed to be related to a financial fraud; it included details of the victims and the sums of money stolen.
Once we analyzed all the available data, it was clear that the C2 was the server-side portion of a banking Trojan infrastructure. We believe the fraud was being perpetrated using Man-in-the-Browser techniques and was also capable of performing automatic transactions to pre-set money mule accounts.
We decided to name this C2 luuuk after the path the administration panel used in the server:/server/adm/luuuk/
Below is a summary of the relevant information extracted from the server side component:
- Around 190 victims, mostly located in Italy and Turkey.
- Fraudulent transactions worth more than 500,000 € (according to logs) .
- Fraudulent transfer descriptions.
- Victims' and mules' IBANs.
The control panel was hosted in the domain uvvya-jqwph.eu, resolving to the IP address 184.108.40.206 during the analysis.
The fraudulent campaign targeted users of a single bank. Even though we were not able to get the malicious code used on the victims, we believe the criminals used a banking Trojan performing Man-in-the-Browser operations to get the credentials of their victims through a malicious web injection. Based on the information available in some of the log files, the malware stole usernames, passwords and OTP codes in real time.
This kind of injections are very common in all the variations of Zeus (Citadel, SpyEye, IceIX, etc.) and all of these are well-known in Italy. During our investigation it was not possible to find the infection vector, however banking Trojans use a variety of methods to infect victims including spam and drive-by downloads.
The attackers used the stolen credentials to check the victim?s balance and perform several malicious transactions automatically, probably operating in the background of a legitimate banking session. That would be consistent with one of the malicious artifacts (a VNC server) we found binded to the malicious server used by the attackers.
Despite the "usual" techniques implemented to steal the users' money (user/password/OTP bypass) what is really interesting in this campaign is the classification of the predefined money mules used to transfer the stolen money.
According to the transaction logs, there were 4 different money-mule (or drop) groups:
- 13test: The limit that the drops in this group can accept is between 40,000 and 50,000 Euros, although there are some drops that have different limits, between 20,000 and 30,000.
- 14test: The limit that the drops in this group can accept is between 15,000 and 20,000 Euros, although there are some drops in this group that have different limits, between 45,000 and 50,000.
- 14smallings: The limit that the drops in this group can accept is between 2,500 and 3,000 Euros.
- 16smallings: The limit that the drops in this group can accept is between 1,750 and 2,000 Euros, although there are drops in this group that can accept a quantity between 2,500 and 3,000 Euros (as in the group 14smallings).
This could be an indicator of a well-organized mule infrastructure. Different groups have different limits on the money that can be transferred to its mules, an indicator of the levels of trust between them.
The operators of this control panel removed all the sensitive components on January 22nd, two days after our investigation started. Based on the transaction activity we believe that this could be an infrastructure change rather than a complete shutdown of the operation.
In addition, based on the fraudulent transaction activity detected in the server and several additional indicators, we believe that the criminals behind the operation are very active. Also they have shown proactive operational security activities, changing tactics and cleaning traces when discovered.
Kaspersky Lab is maintaining contacts with different LEAs and the affected financial institution in order to prosecute the criminals.Kaspersky Fraud Prevention vs. the Luuuk
The evidence uncovered by Kaspersky Lab's experts indicates that the campaign was most probably organized by professional criminals. However, the malicious tools they used to steal money can be countered effectively by security technologies. For instance, Kaspersky Lab has developed Kaspersky Fraud Prevention - a multi-tier platform to help financial organizations protect their clients from online financial fraud. The platform includes components that safeguard client devices from many types of attacks, including Man-in-the-Browser attacks, as well as tools that can help companies detect and block fraudulent transactions.UPDATE
After the publication of the post, our colleagues at Fox-IT InTELL sent us some potentially related information regarding this campaign. According to this new information, the Luuuk server could be related to the ZeusP2P (aka Murofet) infrastructure as we originally suspected.
We received two decrypted configuration files belonging to the ZeusP2P with a reference to the same server where Luuuk was hosted:
The configuration belongs to a botnet named "it" (for Italy). The Luuuk server is being used to host the code that is injected in the victims´ browser. It also manages the automatic transfers to a predefined set of money mules (drops) accounts.
We were also able to analyze the binaries using these configurations. The first one (c8a3657ea19ec43dcb569772308a6c2f) is a ZeusP2P (Murofet) sample that was first seen back in August 2013, months before the malicious transactions were made. It tries to connect to several of the sinkholed servers used to take down GameOver.
This additional data reinforces the theory that the Zeus family is behind the Luuuk server - in this particular case it appears to be of the ZeusP2P flavor. However, this is not definitive proof that the malicious transactions in the campaign were performed by this family, as the injected code on the server was not there when we analyzed it.
We would like to thank Fox-IT for sharing this information.
More than a year has passed since the release of our last article on HackingTeam, the Italian company that develops a "legal" spyware tool known as Remote Control System, or short, RCS. In the meantime a lot has been happened, so it's time for an update on all our current research findings on the RCS malware.Locating the command servers
One of the most important things we've uncovered during our long and extensive research is a specific feature than can be used to fingerprint the RCS command servers (C2s). We presented details of this method at the Virus Bulletin 2013 conference.
To summarize, when a special request is sent to a "harmless" HackingTeam RCS C&C server, the RCS C&C responds with the following error message:
First of all, the codename 'RCS' is there, all right. What we weren't sure about was the 'Collector' referred to in the response. This probably refers to the fact that the server "collects" information from the victims. We used this particular fingerprinting method to scan the entire IPv4 space, which allowed us to find all the IP addresses of the RCS C2s around the world and plot them nicely to a map showing their locations. šWe pinpointed a grand total of 326 C2s.Count of C2s Country name 64 UNITED STATES 49 KAZAKHSTAN 35 ECUADOR 32 UNITED KINGDOM 24 CANADA 15 CHINA 12 COLOMBIA 7 POLAND 7 NEW ZEALAND 6 PERU 6 INDONESIA 6 BRAZIL 6 BOLIVIA 6 ARGENTINA 5 RUSSIAN FEDERATION 5 INDIA 4 HONG KONG 4 AUSTRALIA 3 SPAIN 2 SAUDI ARABIA 2 MALAYSIA 2 ITALY 2 GERMANY 2 FRANCE 2 EGYPT 1 UKRAINE 1 THAILAND 1 SWEDEN 1 SINGAPORE 1 ROMANIA 1 PARAGUAY 1 MOROCCO 1 LITHUANIA 1 KENYA 1 JAPAN 1 IRELAND 1 HUNGARY 1 DENMARK 1 CZECH REPUBLIC 1 CYPRUS 1 Other 1 BELGIUM 1 AZERBAIJAN
The largest amount of identified servers was in the US, Kazakhstan and Ecuador. Unfortunately, we can’t be sure that the servers in a certain country are used by that specific country’s LEAs; however, it would make sense for LEAs to put their C&Cs in their own countries in order to avoid cross-border legal problems and the seizure of servers. Nevertheless, several IPs were identified as “government” related based on their WHOIS information and they provide a good indication of who owns them.Mobile modules
It was a well-known fact for quite some time that HackingTeam products included malware for mobile phones. However, these were rarely seen. In particular, the Android and iOS Trojans have never been identified before and represented one of the remaining blank spots in the story. Earlier this year, we discovered a number of mobile malware modules coming from HackingTeam for the following platforms:
- Windows Mobile
All these modules are controlled by the same configuration type, which is a good indication that they are related and belong to the same product family.
Certainly, our main interest during the analysis of the mobile modules was in iOS and Android, due to their popularity. The iOS module works only on jailbroken devices. Here is a description of the main functionality of the iOS module:
- Control of Wi-Fi, GPS, GPRS
- Recording voice
- E-mail, SMS, MMS
- Listing files
- Visited URLs
- Cached web pages
- Address book
- Call history
- List of apps
- SIM change
- Live microphone
- Camera shots
- Support chats, WhatsApp, Skype, Viber
- Log keystrokes from all apps and screens via libinjection
The Android module is protected by the DexGuard optimizer/obfuscator and is therefore extremely difficult to analyze. However, we discovered (see the trace below) that the sample has all the functionality of the iOS module listed above - plus support for hijacking information from the following applications:
Another aspect of particular interest to us was the way the malware samples are installed on mobile devices. We discovered several modules that infect mobile devices connected to infected Windows or Mac OS X computers.
As already mentioned, the iOS module can only be used on jailbroken devices. That is why the iOS infector uses the AFP2 protocol to transfer. The "infector" has a nice GUI that enables installation if there is physical access to the victim's device or remote admin access to an infected computer.
List of Apple devices supported by the iOS infector
After successfully connecting, the iOS infector copies several files to iOS and runs an install.sh file:
As mentioned above, remote admin access to an infected computer is one of the possible ways for the malware to be installed on a connected mobile device. The fact that only jailbroken iOS devices are supported can be a limiting factor. However, this is not a huge problem since an attacker can also run a jailbreaking tool such as Evasi0n via the same infected computer. In this case the only thing that can protect a user from a remote jailbreak and infection is the mobile device’s passcode. However, if the device is unlocked while connected to the infected computer, it can be infected by the attacker.
Another interesting mobile infector is the one for BlackBerry devices, which uses the JavaLoader application to load malware samples on BB 4.5 and 5.0. In its disassembled code, we found a path to the PDB debug file, which appears to have been mistakenly forgotten by the authors. The original project was located in the ‘C:\HT\RCSBlackBerry\Workspace\RCS_BB_Infection_Agent\’ when this malware was created.
In this latest installment of our ongoing research, we uncovered a huge infrastructure that is used to control the RCS malware implants. Our latest research has indentified mobile modules that work on all well-known mobile platforms, including as Android and iOS. These modules are installed using infectors - special executables for either Windows or Macs that run on already infected computers. They translate into complete control over the environment in and near a victim’s computer. Secretly activating the microphone and taking regular camera shots provides constant surveillance of the target - which is much more powerful than traditional cloak and dagger operations.
The new data we are publishing on HackingTeam’s RCS is extremely important because it shows the level of sophistication and scale of these surveillance tools. We like to think that if we’re able to protect our customers from such advanced threats, then we’ll sure have no trouble with lesser, more common threats like those posed by cybercriminals.Appendix:
MD5s of mobile infectors:
- 14b03ada92dd81d6ce57f43889810087 - BlackBerry infector
- 35c4f9f242aae60edbd1fe150bc952d5 - iOS infector
MD5s of Android samples:
MD5s of Windows samples:
List of active C2s on 19.06.2014:
RCS modules (using Kaspersky Lab’s classification names):