A new variant of the Mirai IoT malware was spotted in the wild when it launched a 54-hour DDoS attack against an unnamed U.S. college.
While the attack occurred on February 28, Imperva Incapsula is informing the world about it today. The researchers believe it is a new variant of Mirai, one that is “more adept at launching application layer assaults.”
The average traffic flow was 30,000 requests per second (RPS) and peaked at about 37,000 RPS, which the DDoS mitigation firm said was the most it has seen out of any Mirai botnet so far. “In total, the attack generated over 2.8 billion requests.”
During the 54-hour DDoS attack on the college, researchers observed a pool of attacking devices normally associated with Mirai such as CCTV cameras, DVRs and routers. Attack traffic originated from 9,793 IPs worldwide, but 70% of the botnet traffic came from 10 countries.
The U.S. topped the list by having 18.4 percent of the botnet IPs. Israel was next with 11.3 percent, followed by Taiwan with 10.8 percent. The remaining seven countries of the top 10 were India with 8.7 percent, Turkey with 6 percent, Russia with 3.8 percent, Italy and Mexico both with 3.2 percent, Colombia with 3 percent and Bulgaria with 2.2 percent of the botnet traffic.
Other signature factors such as header order and header values also helped the researchers identify the attack as a Mirai-powered botnet, yet the DDoS bots hid behind different user-agents than the five hardcoded in the default Mirai version; it used 30 user-agent variants. Incapsula said, “This–and the size of the attack itself–led us to believe that we might be dealing with a new variant, which was modified to launch more elaborate application layer attacks.”
Less than a day after the 54-hour hour attack on the college ended, another was launched which lasted for an hour and half; during the second attack, the average traffic flow was 15,000 RPS.
90% of application layer attacks last less than six hours, Incapsula said, so “an attack of this duration stands in a league of its own.” The researchers said they “expect to see several more bursts before the offender(s) finally give up on their efforts.”
Cerber ransomware variant evades machine learning
Elsewhere, Trend Micro also has bad news in the form of a new Cerber ransomware variant. Cerber has “adopted a new technique to make itself harder to detect: it is now using a new loader that appears to be designed to evade detection by machine learning solutions.”
The newest Cerber variant is still being delivered via phishing emails, but those emails now include a link to Dropbox which downloads and self-extracts the payload. If the loader detects it is running in a virtual machine, in a sandbox, or if certain analysis tools or anti-virus are running, then the malware stops running.
Cerber stops, Trend Micro said, if it detects any of the following are running: msconfig, sandboxes, regedit, Task Manager, virtual machines, Wireshark, or if security products from the vendors 360, AVG, Bitdefender, Dr. Web, Kaspersky, Norton or Trend Micro are running.
Trend Micro explained:
Self-extracting files and simple, straightforward files could pose a problem for static machine learning file detection. All self-extracting files may look similar by structure, regardless of the content. Unpacked binaries with limited features may not look malicious either. In other words, the way Cerber is packaged could be said to be designed to evade machine learning file detection. For every new malware detection technique, an equivalent evasion technique is created out of necessity.
The newest Cerber may evade machine learning, but researchers noted, “Solutions that rely on a variety of techniques, and are not overly reliant on machine learning, can still protect customers against these threats.”