Security Week 08: SHA-1 exactly everything, vulnerabilities in TP-Link routers, cross-platform botnet with Mirai code

British scientists have proven that the SHA-1 cryptographic hashing algorithm has definitely come to an end. Somehow it turned out that this story seems to have devoted the greatest number of references in digests, possibly due to the fact that no one is joking with cryptography: it either works or it doesn't. Let me remind you what it is about: at the end of 2015, a team of researchers from universities in the Netherlands, Singapore and France published a report in which they shared new ideas for optimizing the collision search algorithm using SHA-1. By the estimate, the real attack could have been done in about 49 days, spending about $ 75,000 on cloud computing power.

A collision is when two different objects have one hash. If the SHA-1 algorithm is used to identify an object, then it becomes possible to “slip” another object so that “according to the documents” it will be identical to the original. And it’s not even about hacking encrypted correspondence, although SHA-1 is still quite actively used in cryptography. “Objects” can be documents, certificates for identification of a specific server: the substitution in this case opens up a wide scope for cyber attacks.

But this space is theoretical: to prove vulnerability in practice is expensive. This week, a team of researchers from Google and the Dutch CWI Institute reported that they did, yes, they could ( news , minisite of the project ).

The result of a practical attack on SHA-1 was the creation of two different documents in PDF format, in which the hash coincides (proof - 1 , 2 ). Creating a document required a huge amount of computation — more than 9 quintillion operations or 6,500 years of processor time (if this spherical processor will work alone, of course).
')

The researchers do not disclose the collision search algorithm. Interestingly, Google decided to treat this problem like any other vulnerability: after 90 days they promise to post the source code of the attack. Use the code in practice, of course, can only very wealthy gentlemen. Although it is already known that the applied method is 100 thousand times faster than a brute-force attack, a second approach to the projectile will cost, in the opinion of sofa analysts, half a million dollars in Amazon prices. If you suffer and use only off-peak processor power available at a discount, you get about 110 thousand dollars - in general, it looks like the $ 75k predicted a year and a half ago.

The real collision attack is known in one piece: Cyber ​​spy campaign Flame used this technique to redirect the victim to a fake website when requesting the Windows Update service. But in that case, the discussion was about the MD5 algorithm, which “breaks” in seconds on any equipment (but the practical attack still requires a lot of processing power). With SHA-1, everything is somewhat more complicated, but after practical implementation, nothing interesting will happen. In the near future, they will find an attack that does use SHA-1 as one of the many ways of hacking; this will briefly attract the attention of the media and the community And in five years time, the everyday work of administrators, who are left alone with obsolete infrastructure, will come, which is a pity to throw out, and it is dangerous to use, and there is nothing to replace.

Google on the project page does not forget to remind that the Chrome browser marks connections using SHA-1 as unsafe. This, of course, is great, but if problems arise, it is most likely not in the browser.

We are waiting for ebuild patches for serious vulnerabilities in TP-Link routers
News Details in the researcher's githaba.

Pierre Kim, a researcher from Côte d'Ivoire (!), Reported multiple vulnerabilities in TP-Link C2 and C20i routers (also sold here), among them is a serious hole in the web interface that allows to execute arbitrary code on the router . The RCE vulnerability is exploited quite simply: you need to log into the web interface, go to the diagnostics page and execute any command there, including, for example, you can run telnetd. In future, you can control the router via telnet without any authorization at all.



As usual, in this story it’s not interesting vulnerabilities (they are not so terrible - you still need a login for the web interface, but this is avoidable), but a vendor fixing process. To its credit, TP-Link, a patch for vulnerabilities is about to be released. But initially, the researcher was told in technical support that they simply do not have a contact for information about the holes in the software. The whole process, from detection to (hopefully) the patch, took half a year from the researcher.

Researchers at the Laboratory discovered a Windows botnet attacking IoT devices with the Mirai code
News Research

Well, as found. A botnet organized by someone who speaks Chinese has been around for a long time, but relatively recently he began using the infamous Mirai attack code on vulnerable Linux devices. At the same time, the botnet itself is built from infected Windows systems. The study of experts "Laboratories" provides pictorial examples of the collection of a malicious campaign, which is called, with the world on a string. The attackers blinded the botnet from what it was: they use attack methods on MS SQL servers, the malicious code is signed by certificates stolen from a number of Chinese companies, keyloggers are embedded on the affected systems, all this is built almost on batch files.

In general, it is such an antipode of dangerous, but beautiful, damn, attacks, which proves that the hacking technique of someone else's systems does not have to be perfect in order to be effective. And so come down. Becoming a victim of threats from shit and sticks of this type is probably doubly insulting. Mirai was hooked to a common cause, apparently, according to the same principles: somewhere talnet-client and scripts for mass (or IP lists from the command server) were attacked from the side of Linux-systems bombing with security holes. Cherry on the cake: the malicious code is hidden in pictures stored on the command server.

Antiquities

"Adolf-475"

Resident dangerous virus, is standardly recorded in COM and OVL files when they are loaded into memory. Places its resident copy in the interrupt vector table at 0000: 0020. Contains text: "Adolf Hitler". With probability 1/8 blocks the deletion of files. Intercepts int 21h.

Quote from the book "Computer viruses in MS-DOS" Eugene Kaspersky. 1992 Page 58.

Disclaimer: This column reflects only the personal opinion of its author. It may coincide with the position of Kaspersky Lab, or it may not coincide. Then how lucky.