This article is a translation of the Dutch article Nederlandse tapkamers niet kosjer that will appear in c't 2003-01, which will be available in the shops on december 19 2002

 

Paul Wouters, Patrick Smits

Dutch tapping room not kosher

According to anonymous sources within the Dutch intelligence community, all tapping equipment of the Dutch intelligence services and half the tapping equipment of the national police force, is insecure and is leaking information to Israel. How difficult is it to make a back-door in the Dutch Transport of Intercepted IP Traffic[1] system?

The discussion focusses on the tapping installations for telephony and internet delivered to the government in the last few years by the Israeli company Verint[2].


 

This company was called Comverse-Infosys[3] until half a year ago, but was quickly renamed when the FBI started several investigations against it and arrested some of its employees in the US on suspicion of espionage. (See pulled FoxNews stories, Politech, Cryptome or Google).

People within the Dutch government got worried too. Especially because they had been warned as early as 1998 about the possible back-doors in the tapping equipment. The ex-ministers of interior ("Binnenlandse Zaken"), Peper and de Vries, could not comment. The minister of Justice at the time, Korthals Altes, was asked to report to parliament in december 2001, where he stated that the security measures meet the required level and that an investigation would be started if this, after all, was not the case. No investigation followed.

In april 2002, Kolkert, procecutor in-chief of the Court of Appeals in Den Bosch, demanded clarification in a letter sent to Stein, the state procecutor ("landelijk officier van justitie") and responsible for interception matters. Stein stated that there are no problems.

On august 24 the project leader of the National Interception Organisation ("Landelijk Interceptie Orgaan", LIO) J.Steeg announced that he plans to check the tapping rooms for backdoors. However, when the equipmenent was bought from the Israelis, it was agreed that no one except Comverse personel was authorized to touch the systems, according to the insider of the AIVD (formerly BVD), the Dutch intelligence organisation that spoke to the EO radioprogram De Ochtenden[4]. Source code would never be available to anyone.

Finally, on October 10th, the Council of Chiefs of Police ("raad van hoofdcommisarissen") sent a confidential letter to the vendors of tapping equipment for ISPs and telcos expressing its concern about the situation in the US.

All of this came after questions were raised publicly in the trial against Baybasin, co-founder of the Kurd parliament in excile, about the possible leaks in the Dutch tapping room as well as manipulation of the collected evidence[4b]. Baybasin was recently sentenced to life-long imprisonment for his connections to assasinations, kidnappings and heroine transports. His lawyers called in experts to question them about the possibility that Israel had laid hands on information tapped by the Dutch. The lawyers claim that Israeli then forwarded the information to the Turkish secret service[5]. Baybasin recently told the media about the Turkish government's involvement with crime syndicates.

c't magazine warned about the blackbox problem in its June 2001 issue[6]. Opentap[7] gave similar warnings on the hacker conference HAL2001[8] in august of 2001 and at the Chaos Computer Club (CCC)[9] in december 2001 with a presentation on lawful interception in the Netherlands[10].

Hebrew as crypto

The insiders at the AIVD and the tapping room were interviewed by the radio program of the EO[11]. According to them, the Dutch government and Comverse have a gentlemen's agreement that the Dutch government would get the Comverse systems for a very reduced price and in exchange the Israelis would get full access to all tapped information. The systems still ended up being more expensive than rejected competitors' quotes. The Comverse maintanance contract alone apparently costs more then the installation itself, according to the anonymous sources. Since the leaks seem to be disguised as maintenance, one could say that the Dutch government is actually paying the cost of foreign intelligence against the Dutch state.

Israeli Comverse employees apparently show up in the tapping rooms on a very regular basis for maintanance, since no Dutch are allowed to touch the equipment. The radio program further stated that the maintanance is done using their own Hebrew keyboards and language. They leave the tapping rooms with filled MO-discs and no-one from the Dutch government has any idea what the Comverse people are doing. To make things yet worse, Comverse can dial-in to the tapping room equipment at all times.

The possible criminal nature of Comverse and their overpricing are not the only problems. A comparison of the Comverse tapping records with billing records of KPN, the largest Dutch telco, shows that 20% of the calls that should be tapped, are not tapped at all. The Dutch government still keeps buying Comverse equipment.

On november 26, a day after the EO radio program was broadcasted, three political parties, D'66, GroenLinks and SP asked questions to the government in parliamant. The current minister of interior, Remkens, answered that the chance of the tapping rooms leaking information is small, but not zero. He further claimed that the Comverse employees were given the most strict screening by the Dutch intelligence agency AIVD, and that they are never allowed to work without supervision. Comverse was chosen based on its price-performance results, the minister said.

Hacking the system?

In an interview with 2Vandaag[12], a daily Dutch television news program, defense specialist and LPF party leader Herben believes that there is enough cryptography know-how available in the Netherlands to hack the systems, if Comverse does not assist in the evaluation process. Apparently, Herben hasn't thought about the intrusion detection system that has undoubtedly been installed in these tapping systems by Comverse. He also seems to forget (as did Remkes) that these systems work in Hebrew. On top of it, proving the inner workings of the machines to be correct and safe is anything but a trivial task.

The capacity of the MO-discs and the bandwidth of the dial-up facilities is not enough to copy a lot of internet traffic or entire telephone conversations. A Comverse employee would have to swap disks so often, that he would have to use the tapping room as a hotel. So, assuming that there is no (illegal) high-speed internet connection between the tapping room and the Israeli embassy, what the Comvers staff can do at the most on these visits is to copy a list of who talks to whom, and the cryptographic keys that are used to secure the tapping communications. Therefore, the Israelis don't need to copy entire phone conversations or all internet traffic of a user from within the tapping room, but can simply monitor the encrypted traffic that is sent to the tapping room. Having the cryptographic key to the data, they then decrypt it at their leisure. If any nation has the technical skills and knowledge to pull this off, it is Israel.

The experts

We explained the situation to two cryptography experts: Niels Provos[13] of the OpenBSD team and autor of various crypto software such as Outguess[14], a program to detect steganographic content, and Michael Richardson[15] of the FreeSwan Project, the IPsec implementation of Linux. We posed the hypothesis of the insecure tapping room and asked whether it would be possible for the Israelis to get a hold of our taps.

Provos explains that a very important part of strong cryptography is a good random source. Without a proper random generator, or worse, with a intentionally crippled random generator, the resulting ciphertext becomes trivial to break. Even if Comverse would let experts have a look at the source code, if there is one single unknown chip involved with the random generation, such as a hardware accelerator chip, all bets are off. Provos suggests to use only off-the-shelf PC hardware. If you can trust the hardware and you have access to the source code, then it should theoretically be possible to verify the system. This, however, can just not be done without the source code, according to Provos.

One possible undetectable scheme could be to use a set of truly random, but pre-calculated keys. Only those who know the pre-calculated set, Comverse in this case, could break the cipher, which would become a sort of one-time pad for Comverse only. Provos also pointed us to the work of Adam Young en Moti Yung, who have written a few papers on what they dubbed, kleptography[16], the art of secretly stealing the cryptograhic key from the ciphertext stream itself. Their research showed it is impossible for third parties to detect whether any given ciphertext is secretly leaking key material.


 

An overview of TIIT

The Dutch tapping porotocol, Transport of Intercepted IP Traffic[1] is used for the communication between the tapping machine at the ISP, and the Dutch government. The suspect who is using the internet generates IP traffic that is copied by a special sniffer machine, called S1. The S1 then encrypts the traffic with an RC4 (or AES) key supplied (and generated) by the Dutch tapping room, and sends the encrypted traffic to the S2, the ISP's collector machine. The collector sets up an encrypted connection, using SSL or IPsec to the government collector machine, the T2. This will normally happen over the internet itself. The T1 then sends the encrypted information onwards to one more agencies, who all have their own T2 for receiving the encrypted traffic. The T2's have the key to decrypt the gathered data into the original plaintext, as it was captured by the ISP.


  Both the SSL and IPsec protocol, which are part of the encrpytion scheme used by the Dutch tapping specification (TIIT), contain parts where one has to "fill" packets with random data. It is impossible to see whether this data is truly random, or contains a secret message. This means that no-one needs to go to the tapping room to fetch the key material. According to Provos, the keys can just be sneaked into the encrypted tap itself. Richardson agrees with this view. There has even been a software implementation of this in the past. The TIS-client implemented this feature as "Government Access to Session Keys method". There are even rumors that the ciphers SHA1 and DSS, both NSA ciphers, leak key information on purpose, with only the NSA knowing how to retrieve it.

Richardson claims that it is easy to use weak key material. And there are other dangers as well. Because RC4 is based on XOR, using the same key twice is enough to crack the code. RC4 is used for the inner encryption of user data in the TIIT, since the final AES candidate wasn't known at the time when the protocol was set. But this RC4 encrpytion is packaged in another layer of encryption, SSL or IPsec. That layer needs to be broken as well.

Richardson takes IPsec as example. Imagine that we need to leak an RC4 key and an IPsec key. For RC4, only the first 128bits are relevant. For IPsec 3DES is often used, which means another two times 56bits. Each IPsec packet has an IV of 64 bits. This IV is random filling to ensure that there will never be two identical packets encrypted with the same key, a deadly sin in the world of cryptography. So this makes it possible to hide 64bits in each IPsec packet. Theoretically, after two packets you have leaked the RC4 key, and after another two you have the 3DES key too, although Richardson says that if such a scheme is used, it is very likely that the leaking would take place a bit slower, so it can be covered up. For example, the 64 bits can be divided in four parts of 16 bits hidden in the first 20 bits of four IV's. 16 bits of actual key material and four bits to poiint to the position of those bits in the key. That means that about 16 IPsec packets are needed to leak the entire key. According to Richardson, that would leave plenty of randomness in the IV to make this leakage invisible.

Due to the overhead of IPsec and of the TIIT, this means the tapped user needs to cause even less packets for this to happen. In other words, reading a few lines of email or looking at a single wegpage, is more then enough to leak all key information.

Weis and Lucks showed that the use of the IV isn't even needed, and presented their paper All your keybits...[17] at SANE2002[18] that mathematicly proves that blackbox crpytography is fundamentally insecure and that leaking key material cannot be detected in any way.

Conclusion

Without the cooperation of Comverse, is it not possible to determine whether the Dutch tapping systems contain backdoors or not. Worse, even if Comverse would appear to cooperate, there is no way to detect a possible double-cross. Key information can leak quickly and undetectably and the only way to prevent that is by having full control over both the hardware and the software involved.

In mid december, the parliament will discuss the annual report of the AIVD, but it seems unlikely that the public will ever find out what really happened. Remkes only wants to talk about these matters behind closed doors. De Graaf, party leader of D'66, said he finds the risk of possible manipulation of the tapping rooms "pretty serious", but cannot give more public statements, since he was a member of the watchdog commision that oversees the intelligence service AIVD, and therefore has inside sensitive knowledge.

Remkes claims he didn't know about the dangers. Apparently, he was the last one that didn't know; Comverse and blackbox cryptography have been under heavy fire for years.

Literature

Media links