RFID Passport Security "Poorly Conceived" 33
tonk writes, "European expert researchers on identity and identity management summarize their findings from an analysis of passports with RFID and biometrics — Machine Readable Travel Documents or MRTDs — and recommend corrective measures that 'need to be adopted by stakeholders in governments and industry to ameliorate outstanding issues... By failing to implement an appropriate security architecture, European governments have effectively forced citizens to adopt new international MTRDs which dramatically decrease their security and privacy and increases risk of identity theft. Simply put, the current implementation of the European passport utilizes technologies and standards that are poorly conceived for its purpose.' The European experts therefore come to similar conclusions as the Data Privacy and Integrity Advisory Committee of the US Department of Homeland Security in a draft report, which seems to be delayed."
Why make them unnecessarily data rich? (Score:3, Informative)
The "machine readable" part should not need to contain anything more than a unique number (i.e. primary key) and perhaps a pki type hash to verify authenticity. The rest can be contained in a (hopefully) secure database using an international common format or schema. The id number could also be soley used as a passport serial number and not used for any finanical purposes.
The actual report (Score:4, Informative)
I said it before and I'll say it again... (Score:4, Informative)
microwaves! (Score:4, Informative)
Data Rich == Enhanced Privacy (Score:5, Informative)
In response to the poster who asked why these passports are data rich: Because it avoids the need to place all of this detailed personal information in central databases which are accessed remotely from thousands of locations around the world. How would you secure such a database?
The ICAO recommended approach is much more secure -- the problem here is that the EU has chosen not to implement the security features. The US State Dept. started down the same path, but changed course in response to public outcry.
Here's a description of how the "basic authentication" as recommended by the ICAO specifications works -- this is from memory, but it should be very close to accurate:
So, unless you can break AES or exploit some other flaw in the passport chip* the only way to retrieve the data from the chip is to look inside the passport. If you can look inside the passport, however, you really don't need to talk to the chip at all, because with the exception of some digital signatures, all of the data in the chip is printed in the passport.
What exactly is in the chip? Again from memory:
In the future, other biometrics may be added as well, like a fingerprint image.
The US State Dept. has chosen to go one step beyond the ICAO recommendations and add shielding to the passport cover, so the chip is isolated and can't be queried or detected when the cover is closed. Without that, an attacker couldn't read the data from the chip, but he could "ping" the chip and notice its presence.
*Note that these chips were not created for passports, they're standard contactless smart card chips which have decades of use as security devices behind them, and which protect billions in credit card transactions annually -- nothing's perfect, but they're darned good, having gone through many years of breaks and application of countermeasures.
here's a thought experiment... (Score:3, Informative)
a) 'Sounds interesting, but lets get some more input and make sure there is no downside for our employers, the public'.
b) 'Woot! More power and influence for me! Promotion for you, but if it goes wrong, you will get the blame!'
Way to safely use central DB (Score:3, Informative)
There is a solution:
1. ID reader queries chip to obtain nation of origin.
2. ID reader presents a certificate from the owner with the ID of the reader to the nation of origin, requesting permission to read the passport. Nation of origin authenticates request and provides signed packet with reader ID, valid authorization time range, timestamp, and certificate of nation of origin. This approves that the nation of origin recognizes the reader as a valid one for reading the passports.
3. Reader caches #2 to reduce traffic, and presents this packet to the passport. The passport verifies that the ID reader is approved to query the passport by its nation of origin.
4. Passport returns its ID, certificate, and signed permission to query with some expiration date encoded.
5. Reader presents #4 to nation of origin to query its database. This proves that the passport is physically present.
6. Nation of origin returns signed database entry.
Quick objections might be that this sounds like a lot of round trips, but all but one of these trips could be cached (the reader could be given permission to query passports for a day or more at a time). Another objection might be that every reader would need to get permission from every nation, but this is also not the case - there merely needs to be a chain of trust. So, the US could grant France access to its passports, and then France could delegate access to individual readers.
Various pros and cons exist and I think the actual-implemented solution is not a horrible one. I just wanted to show that a central DB doesn't have to be impossible-to-secure.
Re:Data Rich != Enhanced Privacy (Score:3, Informative)
This is not correct. The EU has implemented those security features - Basic Access Control (BAC) especially is a European development, mainly brought into ICAO by German Federal Office for IT Security (BSI). BSI also proposed Extended Access Control (EAC) for additional data such as fingerprints. The study on which the Budapest declaration is based has all this analysed.
The shielding within the cover is not a complete Faraday cage, see RFID Passport Shield Failure Experimental Report [flexilis.com]
The basic problem is, that
If you have access to the MRZ, you can just decrypt the session keys. Successfull brute force attacks on eavesdropped passport-to-reader-communication is already feasible within hours, see ePassport Privacy Attack [riscure.com]. Once the MRZ is known, e.g. when you have to leave your passport in a hotel or after a successful brute force attack, the passport can be 'pinged' e.g. when going through a door and then be used as a trigger for something. Excessive eavesdropping of passport-to-reader communication e.g. at airports allows for later brute forcing and then identity theft.
The Budapest declaration and the study behind it focus in all these issues and take all your points into account. BAC and what is already known on EAC has been analysed. Still the resumee is 'poorly conceived'.
Well, as the US want to store all the data collected from the passports for 50 years, maybe they have an answer to that question?
The problem is not the chips. The problem is the RFID interface, the limited keyspace entropy, the absence of the option to change the key, well, see above.
Another problem with the passports is the use of biometrics in General, which is also covered within the study and the declaration.
The bottom line is: RF interface and biometrical identification do not increase security, but risks. These passports will cost lots of privacy, security, and tax money.
Data Rich == Enhanced Privacy (Score:3, Informative)
Thanks for that correction. That's what I get for reading only the abstract.
According to that report, the shielding is only ineffective if the passport is open. I suggest you keep your passport in a flat sleeve or put a rubber band around it if you carry it in a purse or bag where it could fall open. I carry mine in a flat pocket, so there's no issue.
Yes, the MRZ contains insufficient entropy. This is the one real flaw in the system, and it needs to be fixed. The failure of the MRZ to be sufficiently hard to brute force doesn't support your subject line, though -- placing all of the data in the passport itself rather than in a database does enhance privacy as compared with a widely accessed central database.
Note, though, that to recover the MRZ via brute force, you have to eavesdrop on a legitimate communication, which means that the only place you can obtain the data is at passport control, which significantly limits an attacker's options. An attacker with a directional antenna concealed in a briefcase could probably eavesdrop on enough data to recover a few passports' contents while waiting in line, but that's about it. Obviously, that could be enough to steal the identity of the passport holders which makes this a problem, but if it becomes clear that it is a problem it would not be hard for immigration agencies to add additional shielding around the readers to attenuate and smear the (already very weak) signal beyond recovery.
The better solution is to fix the MRZ, of course, and add ~50 bits of random data to it.
In the case of US passports, this is true only if the passport is open.
Just for clarification: if the passport is open, the MRZ is not required to ping the passport and use it for a trigger. It's quite likely that in most cases just being able to see the ATR (Answer To Reset) string transmitted by the passport during powerup is adequate to identify the nation of origin of the passport, and other characteristics of the signal might even make it possible to identify a specific passport. To be really sure the trigger targets a specific individual, it would be helpful to have the MRZ, though.
Hmm. It occurs to me that if you can plant an explosive in an area travelled only by people with passports, the *lack* of a response from the passport might someday be sufficient to identify the holder as an American. Hopefully other nations will adopt the shielding.
There's a very large difference between securing a database that is accessible from a few dozen locations in one country and a database that is accessible from hundreds (thousands?) of locations in dozens of countries around the world.
Were the key stronger, I don't think that would be a significant risk. Sure, the key might be brute forceable in less than 10 years, but the value of the data in question is low enough that it wouldn't be worth the effort. It would be much easier to suborn an immigration official to provide the information. Actually, even with the weak MRZ, it's arguably still easier to bribe a few officials rather than bother with carrying a hidden receiver and self-aiming directional antenna.
One common flaw in technical security analyses is that they fail to take into account the whole gamut of non-technical attacks possible.