Arbor Networks reports improved attacks being added; about 500k "old" bots still active.

Mirai—the malware responsible for creating a massive "botnet" of hacked Internet-connected cameras, digital video recorders, and other devices that interrupted Internet services for many last week—is still in action, according to data from the network security company Arbor Networks. An ever-shifting army of about 500,000 compromised Internet of Things (IoT) devices is still being controlled by Mirai, based on Arbor's tracking of the malware's communications. And multiple command-and-control networks are still directing those devices to attack websites and service providers across the Internet. But as previously predicted, new and improved versions of the Mirai malware—based on the openly-published source code Mirai's alleged author posted on September 30—are now appearing in the " and wreaking additional havoc.

In a blog post, Roland Dobbins, Principal Engineer on Arbor's ASERT Team, noted that "relatively high concentrations of Mirai nodes have been observed in China, Hong Kong, Macau, Vietnam, Taiwan, South Korea, Thailand, Indonesia, Brazil, and Spain." Devices that are vulnerable to Mirai takeover, he noted, "are typically listening for inbound telnet access on TCP [port] 23 and TCP [port] 2323," and compromised devices communicate via "a remote-control backdoor" that is also present in Mirai, "accessible via TCP/103." Mirai botnets constantly scan the entire Internet for vulnerable devices, so even when a device is rebooted or reset, it can be compromised all over again within 10 minutes.

Dobbins also noted that "multiple threat actor groups are actively working to expand and improve" the attacks that were coded into Mirai, and that "some alterations in the DDoS attack capabilities of at least one Mirai-derived botnet have been observed in the wild."

In a Skype call with Ars, Dobbins said, "It's a minor enhancement to one of the existing [Mirai] attacks." He couldn't give detail about the enhancements, but he added that "multiple groups are working to enhance and customize Mirai."

The original Mirai code is capable of a variety of attacks against DNS services and websites, in addition to more generic network "flood" attacks based on the TCP, UDP, and Generic Routing Encapsulation protocols. Mirai accounted for most of the attack on Dyn's DNS service on October 21, and was part of earlier attacks on security reporter Brian Krebs' site and on French cloud provider OVH. Those attacks measured over 600 gigabits per second and over 1.5 terabits per second at their peaks, respectively.

While the total volume of traffic thrown at Dyn hasn’t yet been publicly released, Level 3 Communications chief security officer Dale Drew said in a Twitter conversation that the numbers had been shared with major network operators. Drew told Ars in a separate conversation that "tens of millions" of distinct devices were involved in the Dyn DoS attack, and that some of them were clearly not Mirai-infected devices; not all of the devices were necessarily active at the same time.

Dobbins wrote that the "potential collateral impact of DDoS attacks launched by the Mirai botnet can be highly significant." The outbound traffic from hacked devices—including attacks against intended targets and scanning for other vulnerable devices—could crimp the network bandwidth of even major broadband ISPs, causing outages for customers.

Given the wide availability of the code, it's fairly certain that even more Mirai variants will emerge—and make their presence felt as the holidays approach. Criminal botnet operators will likely use Mirai's success as a way to extract blackmail payments from online retailers and banks with threats of interfering with online shopping. Stopping (or at least reducing) those attacks will require network operators to work to identify vulnerable or hacked devices themselves and block the command-and-control traffic to them.