CCC 25C3

2009-02-01

Morton Swimmer

Trend Micro, USA
Editor: Helen Martin

Abstract

Upon leaving the 25th Chaos Communication Congress, Morton Swimmer concluded that 2009 is going to be a very interesting year – but not in a good way. Here he provides a full round up of the research presented at the event.


The Chaos Computer Club (CCC) is a German-based hacking and technology activist group. 25C3, which took place 27–30 December, was the 25th Chaos Communication Congress and my 18th year of attending the event. The congress recently expanded from three to four days, but this year it reached the capacity of its current venue, the Berlin Congress Center. The question now is how the congress will manage its expansion given that, for privacy reasons, tickets are not available for pre-purchase. Although organized by the CCC, other clubs and groups also attend, and the congress is also used as a venue for open-source project meetings and similar events.

As a consequence of the event being full this year it was harder than usual to get a seat (or even standing room) at many of the presentations. Helpfully, the organizers stream and record nearly all of the sessions, so with a working network connection you can watch a live presentation from wherever you happen to be, or watch it later by downloading the recording. I have organized this report roughly by theme, starting with mobile, then Internet security, followed by malware-specific issues, finishing with a number of miscellaneous subjects.

Mobile security

The mobile platform continues to intrigue hackers as the technology becomes better and smart phones become more widespread. While the attention devoted to this platform is worrying, its security is being eroded only very slowly and it is impossible to know when the cataclysmic event will occur that will cause the phone vendors and the service providers finally to realize that they need to take action. Having said that, the first presentation in this category did give me some cause for concern.

Rogue phone networks

We have long considered mobile phone systems a fairly closed issue, but this may be changing. Dieter Spaar and Harald Welte demonstrated a GSM base station they had bought on eBay. The pair had been able to reverse engineer the base station well enough to run it as a mini-GSM network. During the talk (and the previous night) it popped up as ‘001 01’ network and the researchers were able to make voice and SMS connections. Reverse engineering the base station and creating a system for connecting to it was certainly a tidy piece of work, but it is not clear to me whether or how quickly this could evolve into the threat of rogue phone networks. Ordinarily, a GSM phone will not connect to an arbitrary network if its home network is available, but in the US network coverage is patchy. A coverage gap could be filled by a rogue network to execute a man-in-the-middle attack or just to collect IMSI and IMEI data.

This isn’t as far-fetched as it may seem – in the distant past receivers were used on analogue phone networks to sniff phone credentials. Until now, obtaining the required hardware and then understanding it well enough to be able to create a GSM network node has been a hurdle, but Spaar and Welte have now demonstrated that the hurdle can be cleared.

They also mentioned in passing a similar project that is attempting to implement GSM protocol on the GNU radio/USRP (Universal Software Radio Peripheral). If this succeeds, it will be much easier to obtain the necessary hardware, as a GNU radio (which is basically a radio receiver implemented heavily in software and capable of receiving all bands) is far more easily obtained than a GSM base station (although still costly at around EUR 1,000). The authors had bought the entire stock of base stations available on eBay at that time and were offering them at cost to interested parties. The GNU radio and base station projects should complement each other well as one deals with the physical layer while the other with the network link layer.

The authors’ base station did not support UMTS, so they have not yet done any work in that field. With the rise of smart phones, UMTS is the protocol that is the most interesting from the Internet security perspective. UMTS does support a much more sound encryption protocol than GSM, and the encryption goes to the telco rather than stopping at the base station, so it will be much harder to find useful hacks. However, as this proof of concept has shown, it is likely only to be a matter of time.

Symbian platform

Collin Mulliner of the Fraunhofer-Gesellschaft Institute presented his findings on Symbian OS 9 security. He chose to focus on this version as previous versions of the OS didn’t have notable security mechanisms in place and the API had become more or less POSIX-compliant with this version. The Symbian OS runs on approximately 50% of the world’s smart phones as the PDA-side operating system. Mulliner was interested in the susceptibility of the OS and its applications to buffer overflow attacks and the usefulness of the capability-based security system.

The OS itself has no buffer overflow protection mechanisms in place, such as stack canaries, but will use the code execution protection mechanisms of the CPU if they exist. For that reason Mulliner found that CPUs with ARM architecture version 5 and below were vulnerable to attack, but version 6 and above were not. He intends to test other types of code injection attack against these newer architectures in the future. He showed the results of some of his experimentation with buffer overflows, such as making a call from shellcode and an IMEI reader. To make shellcode generation easier for this platform he created a shellcode encoder so as to avoid zero bytes and to avoid problems with the instruction cache. Finding exploitable code was done by attaching the (now free) remote debugger and fuzzing. An alternative he mentioned was using IDA Pro’s debugger for Symbian.

Mulliner also discovered what he believes is a weakness in the capabilities system of Symbian OS. By granting network access to virtually all applications, the shellcoder can use networking to take the available IMEI to the open code-signing site that Symbian provides for developers and have code signed specifically for that phone. The captcha system on that site is weak and can be broken by numerous anti-captcha services. Thus an attacker can target a mobile phone via a buffer overflow attack and then create malware for that phone and upload it. Since users are already preconditioned to press the ‘Yes’ button as many times as the phone asks them, he believes the success rate of such malware will be high.

iPhone

This talk was basically a deconstruction of the iPhone’s hardware based on the work the authors did while jail-breaking it (removing the phone’s protection mechanisms). The phone is divided – like most, though not all, smart phones – into two halves: the phone and the PDA part.

The phone part deals only with making calls and transferring data over the GSM/UMTS network. It is based on the Nucleus OS and uses encryption extensively. It has a code trust chain that starts when the initial SIMloader is run, but the bootloader which comes next has exploitable bugs – which the authors found via hardware fuzzing and allowed them to break the trust chain. While no applications can be installed on the phone side of the iPhone, unlocking it is the only way to run the iPhone on networks other than the one for which it was configured.

The PDA part is based on Mac OS X with an XNU kernel. There is a nearly complete chain of code trust, with the bootloader being the exception, so this part is what needs to be broken. The phone is pretty well secured if it isn’t jail-broken, but inevitably someone will find a vulnerability and there is currently no official way of writing software for the kernel of the iPhone (or the phone part), so there is no way of easily protecting the phone with third-party software.

I was a bit disappointed that the authors didn’t go into more detail about the operating system. However, the fact that the OS is based very strongly on Mac OS X is probably enough to understand how it is built. (Later, I was told that the main differences from Mac OS X are in the GUI model, which is far cleaner and tailored to small devices.)

DECT

A group of researchers from Germany and Luxembourg presented their findings on DECT phones. Although used in very few wireless phones in the US, DECT is one of the most popular standards for wireless home phones and other short-range digital wireless devices in Europe and other parts of the world.

Communications are scrambled and presumably time-sliced in an unpredictable manner. Initially, the researchers planned to use a GNU radio/USRP to build a sniffer for DECT wireless traffic, but then stumbled upon a cheaper solution in the form of a PCMCIA card. After reverse engineering that card they were able to create a DECT sniffer for about EUR 23, which was also able to transmit (in contrast to the USRP solution where real-time transmission is hampered by the delay in processing the incoming signals).

Like GSM, the cipher used by DECT is kept secret, but security through obscurity will eventually fail unless the security is very sound. Some parts of the cipher are documented in a patent, but others are not because the encryption is implemented in a dedicated chip, and the authors had to resort to chip reverse engineering.

The end result is that with a fairly small investment in hardware, the researchers were able to eavesdrop on DECT calls, but also sniff traffic from other devices that use this system, such as electronic payment terminals and traffic lights. Eventually they will be able to associate with any given DECT network and place calls.

Internet security

Internet security is always a big topic. This time a few of the talks broached fairly advanced themes.

DNS mess

In July 2008, multiple vendors issued patches to their DNS servers and clients to fix a problem that had been discovered many years ago, but which was only recently recognized as severe. Like much of the Internet, DNS was designed with a threat model more focused on availability than security and is brilliant at distributing data in a federated manner. For that reason, we have come to rely on DNS to federate all sorts of data that probably shouldn’t have been entrusted to a relatively insecure protocol. The bug in question allowed a type of DNS cache poisoning due to the fact that the source ports of the DNS server were predictable.

Dan Kaminsky took us on a tour of all that he had been doing last year getting DNS patched in a coordinated way so that no vendors would prove to be in a liable position. The good news is that it is estimated that over 75% of the DNS servers were patched after one month, but there are still plenty that are not. Achieving this was no easy feat as it required the vendors to be persuaded of the severity of the problem and of the need to fix it pre-emptively. It was a good tale of setting aside egos for the greater good.

Kaminsky put forward a very strong case for getting DNSSEC into a usable state, as he grudgingly believes that it is the only way of making DNS trustworthy in the long run. Dan’s thesis is that authentication on the Internet is basically broken and fixing DNS would go a long way towards re-establishing trust. DNS itself may not, ultimately, be fixable so Dan believes that fixing and then deploying DNSSEC is the only way forward.

Debian RNG

Luciano Bello and Maximiliano Bertacchini reported on a bug that was introduced into the Debian version of the random number generator (RNG) due to a complaint. Apparently, some of the code in the RNG was generating warnings and a user considered this enough of an annoyance to report it as a bug. The code in question read unallocated memory for seeding purposes, so the debugger reported it as an uninitiated variable. Unfortunately, it was not clear to the Debian package maintainers what that code was supposed to do, so they created a Debian-specific patch that removed the two lines in question. The result is that Debian Linux and all Debian-derived distributions had the bug for a while, meaning that a large number of systems are affected.

This affects many cryptographic protocols, in particular key generation and the Diffie-Hellmann key exchange protocol. The problem is that even though Debian has now been fixed and many systems have probably been updated, we may have to live with the bad keys for a long time. Luckily, users of Firefox are not susceptible (even if the web servers it connects to are) as it uses its own crypto library. This vulnerability also highlights how insecure code may be socially engineered into open or closed source software.

MD5 and CA Certs

The most eagerly anticipated event of the conference was the demonstration of an attack against a Certificate Authority (CA). The organizers didn’t announce the title or content of this talk until a few hours before the event, for fear of a court injunction or similar action preventing it from going ahead.

We have known for many years now that the MD5 cryptographic hash function is broken and other hash functions are under attack (see VB, October 2004, p.13). There have been demonstrations of engineered MD5 hash collisions in the past. However, many people still continue to use MD5 in security applications – and, specifically, they are still being used in the generation of cryptographic certificates for websites and code.

Using the known prefix attack against MD5 and deriving the way that Equifax’s RapidSSL service generates website certs, the researchers were able to create an MD5 collision for a CA certificate using the site cert generated for them by RapidSSL. This was a clever piece of engineering as the first fields of a cert are generated by the CA, so they needed a way of predicting these, while the rest of the certificate was under their direct control. The prefix attack allowed them to create a CA cert document for which the signature of the real CA still held true.

With this rogue CA certificate, the researchers could create any sort of site or code certificate they wanted and it would look like the rogue certificate authority was certified by Equifax. They hobbled the generated CA cert by backdating its expiry date to 2004 (the year of the first publication of hash collisions) so that any certificates it generated would not be accepted in practice. While experimenting with this, they found that most browsers do not honour the certificate revocation system, the notable exception being IE in Windows Vista.

The attack is only possible because of the weakness of MD5, and the moral of the story is that we need to stop using MD5 in security applications as soon as possible. The researchers recommended SHA-1, although it too is beleaguered, and so it might be prudent to move beyond SHA-1 and implement a certificate management system.

We had a lively debate about the significance of this presentation. I think it is a significant proof of concept, and even though the researchers picked on RapidSSL, there is a good chance that someone will find a way to abuse a different CA as well. It is always important to have a proof of concept to kick people into action. However, we’ve known for some time that MD5 is broken. We also know that people don’t pay enough attention or don’t understand the certificate system when browsing. Given the DNS and RNG problems, authentication in the Internet is pretty broken at this point anyway. So, it may not make much of a difference. What worried me was that when I met the speakers briefly, my impression was that there is more in the pipeline – they did not reveal everything they knew. It will pay to be vigilant.

The Cisco IOS

Felix Lindner (FX) talked about attacks against the Cisco IOS. Basically, if motivated enough, instead of concentrating on servers, an attacker can take control of the infrastructure instead. Since Cisco is, by a large margin, the market’s biggest supplier of networking equipment, it makes sense to concentrate on Cisco’s IOS operating system. Having said that, the attacker would need to be pretty motivated and it would have to be a part of a targeted attack. The IOS comes in many versions compiled in different ways by different engineers. In fact, OS diversity is probably Cisco’s best defence, as the OS itself is just as susceptible as any piece of software. It would also be a very effective attack and would give the attacker control over the infrastructure. To make matters worse, FX explained how difficult it is to perform post-mortem forensic analysis on IOS: if an error occurs, the equipment reboots itself, removing the evidence of the attack.

To counteract this, FX has been working on tools to log and diagnose the state of the IOS. He also suggests one simple solution: restrict connections to the infrastructure machines, as there is no reason why a user PC needs to access the router directly.

Malware

There were a few talks on malware, but little of any real significance. All I really learned from a talk on SWF malware is that Actionscript 3 will make Flash much more interesting, but it is still rarely used. Future versions of Flash may include P2P and UDP libraries, making it something worth keeping an eye on. Oddly enough, the ‘Curse of Silence’ attack against some S60 phones was shown as a part of the ‘Security Nightmare’ session. Apart from that, most of the talk was tongue-in-cheek speculation.

Storm

While the Storm worm seems to be subsiding a little, it is still a fascinating beast. There was a talk about how to take over parts of the Storm net. This relies heavily on understanding how Overnet works and how it uses hashes. Basically, a Storm bot uses the Overnet to locate the C&C and then communicates directly with it. It turns out that one can poison the hash system and point the bots to one’s own system. One needs to know how to pretend to be the C&C to do this, and the authors gathered their knowledge by reverse engineering the bots and looking at the traffic. They could have ‘cleaned’ all the bots in the network, but refrained from doing so. It probably would be illegal in most countries – and dangerous in any case.

Other talks

A new technology is coming to mobile phones: Near Field Communications (NFC). This is a sort of phone-enabled RFID and is being pushed by telcos but appears only to be available so far in one phone. The suggested uses for the technology include: fare-collection, physical access control, electronic tickets and social networking. To be used in such situations the technology would require really good security, but it looks like it is based on the rather broken Mifare (TM) system with the possibility of using some extra phone CPU power. The work of VU Amsterdam has shown that all sorts of security problems can be caused by bad RF security implementations.

Fabio Yamaguchi talked about vulnerabilities in the TCP protocol that can lead to DoS attacks. His argument is that the assumptions of the attack model made when TCP was designed are different from the current reality. They were more concerned with nodes and network connections being unavailable due to bombings or sabotage than an attack from inside. Apart from existing attacks that we have already seen, Yamaguchi talked about abusing the congestion flow control system to overload the Layer 5 servers or clients.

Tillmann Werner showed off a system that is able to extract signatures directly from the trace data of a honeypot. He had a few neat tricks to make it work smoothly. The system is called nebula (http://nebula.mwcollect.org/), and it refines the signatures as more similar attacks are caught.

Conclusions

As usual, the CCC Congress was very intense and informative. It is an opportunity to exchange ideas and discover new ones in an open environment. With three parallel tracks, workshops and other events on the side and parties at night, there is not much time to track down and talk to people one knows or wishes to get to know. The fact that nearly all of the talks are recorded relieves the pressure of squeezing oneself into the over-full talks to see them live, but it nearly obliges one to review the over 150 hours of video that was created to see what one missed. So, after surviving the four days, I usually spend weeks going over missed presentations while trying to get rid of whatever germs I picked up at the event.

The MD5 event was intended to be the highlight of this year’s Congress, but on its own it failed to impress. However, taken together with the DNS, Debian RNG and some of the other problems that were discussed, a very bleak picture is being painted. It is emerging that the foundations of the Internet are shakier than we thought and the application layers are being developed with little thought for security. My thoughts when leaving the Congress were that 2009 is going to be a very interesting year – but not in a good way.

twitter.png
fb.png
linkedin.png
hackernews.png
reddit.png

 

Latest articles:

Nexus Android banking botnet – compromising C&C panels and dissecting mobile AppInjects

Aditya Sood & Rohit Bansal provide details of a security vulnerability in the Nexus Android botnet C&C panel that was exploited to compromise the C&C panel in order to gather threat intelligence, and present a model of mobile AppInjects.

Cryptojacking on the fly: TeamTNT using NVIDIA drivers to mine cryptocurrency

TeamTNT is known for attacking insecure and vulnerable Kubernetes deployments in order to infiltrate organizations’ dedicated environments and transform them into attack launchpads. In this article Aditya Sood presents a new module introduced by…

Collector-stealer: a Russian origin credential and information extractor

Collector-stealer, a piece of malware of Russian origin, is heavily used on the Internet to exfiltrate sensitive data from end-user systems and store it in its C&C panels. In this article, researchers Aditya K Sood and Rohit Chaturvedi present a 360…

Fighting Fire with Fire

In 1989, Joe Wells encountered his first virus: Jerusalem. He disassembled the virus, and from that moment onward, was intrigued by the properties of these small pieces of self-replicating code. Joe Wells was an expert on computer viruses, was partly…

Run your malicious VBA macros anywhere!

Kurt Natvig wanted to understand whether it’s possible to recompile VBA macros to another language, which could then easily be ‘run’ on any gateway, thus revealing a sample’s true nature in a safe manner. In this article he explains how he recompiled…


Bulletin Archive

We have placed cookies on your device in order to improve the functionality of this site, as outlined in our cookies policy. However, you may delete and block all cookies from this site and your use of the site will be unaffected. By continuing to browse this site, you are agreeing to Virus Bulletin's use of data as outlined in our privacy policy.