Examination of Witnesses (Questions 1-19)
DR GUS
HOSEIN
28 FEBRUARY 2007
Q1 Chairman: Doctor, you are
very welcome.
Dr Hosein: Thank you for having me.
Q2 Chairman:
I am sorry your colleague has been waylaid. Thank you very much
for coming to give evidence to us. For the record, this meeting
is on the record and it is part of this Committee's inquiry into
the EU/US Passenger Name Record Agreement, to be known as PNR
from now on. Dr Hosein, I wonder whether, on your own behalf and
on behalf of your colleague, you would like to give us an opening
statement.
Dr Hosein: Absolutely.
Q3 Chairman: Would you be prepared
to describe not only your own connection but, also, insofar as
you are able to, speak on behalf of Privacy International. I know
that Simon Davies is not here actually speaking on behalf of Privacy
International but I think it would be helpful for the Committee
to have your picture of what both organisations are and your interest
in the subject of this inquiry.
Dr Hosein: As background, today I am speaking
as a Visiting Scholar of the American Civil Liberties Union, which
is the largest civil liberties organisation in the United States,
and which has taken the government to task on most of the anti-terrorism
measures in the past few years. I am also a Senior Fellow of Privacy
International, which is the organisation that Simon Davies is
a Director of. Together, Simon and I have been working on EU/US
surveillance measures since the 1990s, but particularly since
2001 with the advent of passenger data transfers and the more
recent transfer of financial information, and so on and so forth.
Together, Simon and I are also Visiting Fellows of the London
School of Economics where we have done research on this issue
and other issues, such as ID cards and communications surveillance.
We have spoken before at various Committees in the House of Lords
on those issues in the past. Specifically on Passenger Name Records
and the transfer of data to the United States, we denote that
US law does have a number of protections and safeguards; they
have the Privacy Act 1974, for instance, which is a relatively
strong safeguard against the abuse of information. The problem
with that regime, however, is that it only protects the interests
of US persons. That is, if you are not a US citizen or US resident
that law does not apply to you. So that is why the whole transfer
of data between the EU and the US is such a massive legal problem
because American citizens have rights within the United States
but EU citizens do not. Meanwhile, within the EU regime EU citizens
and foreigners have privacy rights, and that is the battle between
the two blocs. I am not sure if you were briefed on the latest
news that has emerged from the United States but the Washington
Post today has run an article about a new data mining programme
that has emerged from the Department of Homeland Security that
seems to be the son of Total Information Awareness, which was
the programme that was shut down a few years ago; somehow it has
re-emerged and the Department of Homeland Security has promised
the American public they have nothing to worry about because they
have only been using foreigners' data and across America there
is not that concern about the data of foreigners. That is the
general problem that we deal with on a daily basis when it comes
to this issue; the legal black hole, and then there is the inattention
to data outside of the US by Americans.
Q4 Chairman: Thank you very much.
You used the expression "data mining". Can you explain
what you mean by that?
Dr Hosein: Data mining is the collection of
vast amounts of information from various sources and then running
algorithms against that data to draw patterns and conclusions.
So if you are able to develop a profile of what you imagine a
terrorist would be like—what kind of mobile phone company
that terrorist would use; what kind of travel patterns, what kind
of telephone patterns, and so on and so forth—and then collect
vast amounts of information about vast amounts of people the theory
is that using advanced algorithms you can identify other people
with similar profiles to that terrorist.
Q5 Lord Foulkes of Cumnock: Can I
ask one question about the organisation? Ironically, I know a
bit more about the American Civil Liberties Union than about Privacy
International. Although Simon Davies is not here you obviously
work very closely with him. Is it a membership organisation, Privacy
International, and how many members does it have, if it is?
Dr Hosein: It is not a membership organisation
it is a watch-over organisation with an advisory board of 50 people
from 30 countries established in 1992. So we have been a London-based
organisation since that time.
Q6 Lord Foulkes of Cumnock: Is it
a registered charity?
Dr Hosein: It is a limited liability company.
Lord Foulkes of Cumnock: It is a limited
liability company. Does it have a sort of charter or—what
do they call these things?
Baroness D'Souza: A mission statement.
Q7 Lord Foulkes of Cumnock: A mission
statement! Thank you very much.
Dr Hosein: Yes, I have not memorised the mission
statement but, generally, we act as a watch-dog organisation looking
at surveillance policies and freedom of expression policies emerging
around the world. We have conducted campaigns in Australia, the
Philippines, New Zealand and, of course, in the United States
and across Europe. In the past few years we have been very busy
with the European Parliament and lobbying against some of the
proposals emerging from the European Commission.
Q8 Lord Harrison: Can I just ask
a supplementary on data mining? Has it existed long enough for
conclusions to be drawn as to how effective it is? If these algorithms
are used against vast amounts of data have they indeed thrown
up profiles of people who are then examined and who are found
to fit the frame and, indeed, may have had such tendencies that
we should want to be aware of? I gather from your tone that you
have some doubt about the effectiveness, but I do not know whether
perhaps you could illumine that.
Dr Hosein: For the sake of being fair and slightly
academic, I will give you one example where it does not work and
one example where you could say it has worked. The jury is still
out as to whether data mining works; there have not been enough
open studies to verify what kind of algorithm was used, whether
it was a fair algorithm, what kind of data was used, whether the
data was actually reliable, which is one of the key problems because
sometimes the data is inaccurate but they then make judgments
against you. This is exactly the problem with PNR, but I will
come back to that in a second. The first case I will give you
is MATRIX. The acronym stands for multi-state access to law enforcement
information, but I cannot remember the exact words. It was a private
company that offered this service to the US Government after September
11 2001. They approached the Department of Justice in the United
States and said: "We are willing to run a data mining office;
set up data that we have gotten from member companies and from
the Government and we are going to help you identify terrorists".
The company also told the US Government that because of the law
in the United States the US Government is unable to do this but
because this company is a private company they are able to process
this information without any due regard for civil liberties. So
the company then ran, against this mass dataset, the profiles
of the 19 hijackers and said: "Who else in the United States
resembles these 19 hijackers?" If I recall correctly the
exact number was 250,000 Americans who were just like those 19
hijackers. So there was a lot of fury but, also, excitement over
the idea of data mining. Apart from selling this product to the
Department of Justice, this company then went to all of the States
to try to sell it to each and every one of the States, so that
their police would be able to have access to these data stores,
and so on and so forth, but because of the public uproar and because
of the number of errors that emerged every State ended up abandoning
that programme and falling out of it to the point where there
was only one State left standing, and that was the State of Florida.
Then, finally, the project came to a close. So that is one of
the cases where it does not particularly work well. In the Washington
Post article this morning I read about this new data mining
scheme that has not been tested openly, but the proponents of
the scheme have said it has proven to be useful in the situation
of Guantanamo Bay, where they have data-mined the detainees. Originally,
before the data mining, they identified the detainees they knew
were caught up with bad people and they developed a profile based
on those bad people and applied it to the rest of the detainees,
and were then able to identify the detainees who were innocent.
So in that sense data mining is applied to prove that you are
an innocent, not to identify the guilty. So that is a case where,
arguably, you could say data mining has worked in that small situation
based on I-do-not-know-what kind of data they have on these detainees.
However, when you apply data mining to a large population, such
as a country, you have the inherent problems of how good is the
algorithm, how reliable is the data and how legal is this entire
process? In the United States, for instance, data mining is tantamount
to being illegal. After the uproar over Total Information Awareness,
which was the first programme to be developed by the Department
of Defense, eventually Congress (and, remember, this is actually
a Republican Congress) pulled the funding and said: "You
are no longer allowed to do research in this area". Then
there was another data mining programme through passenger surveillance
of Computer-Aided Passenger Pre-Screening programme, just to make
sure that violent offenders do not board aeroplanes, and eventually
the funding was pulled on that and the Department of Homeland
Security admitted that the system could not be built, so they
had to move on. On top of that, a piece of legislation was passed
saying that no funding can go towards a data mining programme
of any sort in the United States, ones that apply to the general
population. Somehow these programmes keep on emerging but there
is no funding for them.
Q9 Lord Jopling: Who are the people
who are doing the research to try to demonstrate that data mining
works and what access do they have to what some of us would think
was the most sensitive intelligence information, like, for instance,
material gathered by Menwith Hill in Yorkshire? How deeply are
the people who are trying to establish this system permitted to
go into sensitive intelligence information?
Dr Hosein: To begin with your last question
first, the individuals who are involved have the highest levels
of clearance in the United States Government. They are former
directors of agencies. Two of them have been admirals in the Navy—so
they have the highest possible clearance. On the exact departments
that have been doing this work, one department was the Department
of Homeland Security. They have been responsible for the passenger
surveillance programme, in particular, and that applies particularly
to the sub-agency within the Department of Homeland Security,
which is the Transport Security Administration. Another department
in the Department of Homeland Security that has been doing data
mining has been the Customs and Border Police. That is the department
that, it emerged, was doing profiling of all passengers to the
United States. This news just emerged in November about an Automated
Targeting System. The most significant research programme on data
mining emerged from the Department of Defense, from the Department
of Defense that was responsible for the creation of the Arpanet,
which eventually evolved through to become the Internet. This
is where the brightest of the bright do their research within
the Department of Defense, and they were responsible for Total
Information Awareness (or the Terrorist Information Awareness
programme, as it was renamed). So it is always those with the
highest clearances making use of various forms of data. You rightly
pointed out the data issue.
Q10 Lord Jopling: Before you go on
to that, you did refer to "private company" earlier
on. Does some of this research take place outside the various
defence departments of state, whoever they are?
Dr Hosein: There are a number of private contractors
who are interested in selling these types of services. So the
company behind the MATRIX system was eventually bought up by Reed
Elsevier, the publishing company. There are a number of consultants
who advise on these projects from the private sector, particularly
from Booz Allen Hamilton, which is a company that is the largest
contractor for IT to the US Government, particularly because three
of their five directors are former heads of the intelligence agencies,
such as the National Security Agency and the Central Intelligence
Agency. One of their vice-presidents was appointed the Head of
National Intelligence in the United States. So it is these types
of firms that are aiding the research in this field. There are
also academic institutions, and I am afraid I cannot list them
off the top of my head, who get funding to conduct research on
the algorithms of data mining, but they do not get access to the
wide variety of data. To come back to the data issue, the source
of the data is always a source of controversy. At first the US
Government approached a number of airlines in the United States,
such as Northwest and JetBlue, and there are quiet agreements
to transfer all data to the Department of Homeland Security to
test out these algorithms historically, to see if they would be
able to sustain a real environment. Once the news of these transfers
emerged, passengers filed a number of legal complaints against
these companies, saying: "We did not consent to our data
to be used". When that strategy failed the Department of
Homeland Security then said: "Okay, we will use the data
of foreigners", particularly the EU passengers under the
PNR Agreement. So the PNR Agreement that emerged in 2004 between
the EU and the US permitted for the use of the data on EU passengers
for the testing of what was then known as the Computer-Aided Passenger
Pre-Screening system, or CAPPS-II. The most recent news about
data mining from the United States is that it uses foreigner data
and anonymised data of American citizens, so, again, that is going
to be a source of controversy because American citizens are going
to ask: "How anonymised is this data before you mined it?".
Q11 Baroness D'Souza: I gather from
what you say that the ban on funding for the Government itself
to undertake this kind of data mining activity does not extend
to the American Government paying private companies or academic
research institutions for research that they are doing?
Dr Hosein: I believe that is entirely correct.
Q12 Baroness D'Souza: So, in fact,
the American Government is funding it.
Dr Hosein: Absolutely. After the Total Information
Awareness programme met its demise a lot of the researchers and
the research funds that had been allocated to this were actually
put into what are called "black budgets", so within
the budgets that other people could not see. Then there is the
other strategy, which was when we found out about the Automated
Targeting System, which is the data mining programme that is conducted
at the US border, which we just discovered in November of this
past year. The way that the US Government got around the funding
problem is that this targeting system was originally applied to
cargo. So it was developed and funded to apply to cargo—the
ship moving the cargo, who sent the cargo, and so on and so forth.
What the Customs and Border Police failed to notify Congress was
that they then started adding passenger data to this same machine.
So, in a sense, additional funding was not required. In a sense,
it was a "bait-and-switch"; it is: "Here, fund
this cargo-profiling system", and, "Oh, lo and behold
we are applying this to passenger data". It is leading to
Congressional hearings because there is this concern that regardless
somehow this scheme is illegal under US law.
Q13 Lord Marlesford: Is all of what
you have been saying purely related to terrorism or does it also
cover other forms of serious crime?
Dr Hosein: It is important to note, first, that
the PNR transfer agreement applies to terrorism and serious crime
and organised crime of a trans-national nature. So it is a little
broader than terrorism already. The MATRIX system was being applied
beyond terrorism as well, and the Automated Targeting System (again,
this is something that we learned about in November), according
to its mandate and according to the statements made to the public
from the Department of Homeland Security, is for combating terrorism,
serious crime and border offences. So it has broadened out in
its purpose, and I would almost say that its primary purpose never
was terrorism per se; its primary purpose was to help manage the
flow of information heading towards the US Government regarding
foreign travellers.
Q14 Chairman: I think Lord Marlesford
has very helpfully set the scene for our first question! You have
defined for us very concisely the purpose of the PNR agreement,
which is to enhance the security of the United States, both against
terrorism and against organised crime. To what extent do you think
this has actually been effective? It is a difficult answer, but
can I also, at the risk of pre-empting a later question, ask you
to draw a comparison between the effectiveness of the PNR Agreement
and the effectiveness of a visa system?
Dr Hosein: It is inevitable that the mass surveillance
of a given population has some advantages; it is inevitable that
you will be able to pick out somebody amongst the masses. For
instance, with the US VISIT system, which is the fingerprinting
system at the borders, the government is very fond of saying they
have had 80 million people travel through the United States and
get fingerprinted and they have caught 2,000 people. They did
not say they were 2,000 terrorists but they said they had caught
2,000 people. If you do the calculations, that is a 0.000025 per
cent success rate. So I am not entirely sure what you are looking
for when it comes to effectiveness. It is easy to say how effective
this is, it is harder to prove it, in the sense that when passenger
data was first considered as useful data for the US Government,
it was originally considered for the Computer-Aided Passenger
Pre-Screening System, or CAPPS-II (which I will refer to as CAPPS-II
from now on). Since that time the CAPPS-II system has failed.
The former Secretary of Homeland Security, Tom Ridge, when he
gave his press conference about how they were shutting down CAPPS-II,
he literally took a knife to his heart and said: "We're cutting
it; we're killing it; it's dead", not only because of the
privacy concern but the technological concerns. This is where
we get into the whole area of the actual effectiveness of this
data. I have spoken at length with a number of industry officials
in the airline industry and they always reiterate very carefully
that PNR is not some beautiful dataset that every government is
dying to get their hands on; PNR is a highly unreliable set of
data. There are misspellings of names and there is inaccurate
information taken down generally about eating, seating, and so
on and so forth. It is not this perfect set of data. As a result,
how useful is this imperfect set of data? When you promise to
use it for data mining it is highly dangerous because you are
going to start identifying the wrong people for the wrong crimes,
as we have seen emerge in the United States with their own watch-list—the
"no-fly" list is what it is called—where they
verify names of people before they get on aeroplanes within the
United States. There have been 30,000 complaints against that
watch-list of people who have been unable to get on `planes. We
have not heard similar situations about foreigners being unable
to get in the United States because there is no duty to report
on these issues.
Q15 Chairman: Can I interrupt you
there? You referred to unreliability. To what extent, from your
understanding of where the PNR Agreement stands, does the Agreement
take into account that degree of unreliability?
Dr Hosein: Absolutely not at all. In all the
political rhetoric (and this debate is mostly about political
rhetoric) it has only been about how this data is inherently useful
for the war on terror; there has been no critical questions regarding
the integrity of the data. I have to admit, when I discovered
that this—it was only about a year ago—data was not
reliable, I was shocked because I bought into the whole argument
that this was essential.
Q16 Chairman: I am sorry, I interrupted
you.
Dr Hosein: One final point, which is that it
is important to remember that when the US Government passed the
law regarding access to Passenger Name Records it was actually
just one line within a large piece of legislation. The one line
said that the US Government may demand from foreign carriers to
hand over Passenger Name Records. That is all it said. It is a
massive piece of legislation, and one line. Somehow that one line
emerged and developed into this massive surveillance system. I
do not believe the US Congress is fully aware of how far it has
gone.
Q17 Lord Jopling: The Agreement lists
the 34 "data elements". I wonder if you could tell us
to what extent those 34 items are used. Are there some which have
not been used at all? I wonder if you could tell me which of them
are, in practice, used and exploited.
Dr Hosein: I have to provide a very large caveat
before I answer that, which is that very few people understand
how this data is being used, and the reason very few people understand
is because there has not been open review. The only review of
the US Government's use of Passenger Name Records occurred in
2005 and was conducted by a number of EU officials and two officials
from the Data Protection Commission Offices; one being the Assistant
Information Commissioner in the United Kingdom and one being a
Deputy Commissioner in Germany. They conducted the only review
that has taken place and when they conducted the review they were
forced to sign non-disclosure agreements by the US Government,
so they could not, other than what was within the legal mandate,
discuss the general operations of the data mining and how it is
actually conducted. That report was only published last year (the
report that eventually emerged, which was heavily redacted) and
still to this date we are unable to fully understand how this
information is being used. So, having said that, the controversial
points when it comes to 34 fields of data focus particularly on
what are numbers 26 and number 27, the OSI information and the
SSI/SSR information. These are the fields that could contain sensitive
personal information of a medical nature or a religious nature.
So this could be where a request for halal food is made or if
it says you have heart disease. It also discloses information
about your travelling companions, so it could say you are travelling
with somebody who is not your official, legal partner, and so
on and so forth. The situation after the first agreement between
the EU and the US was that this data would be `pulled' to the
United States by the US Government, which would then log into
the reservation systems and grab this data themselves. Or, if
it was possible, the carriers would send this data to the US Government,
at which point the US Government would filter the sensitive fields.
So they would go through all the data and delete the data that
said anything about the food requests or your travelling companions.
The review that took place in 2005 gave the US Government a favourable
note on this; it said that the US Government had effectively implemented
this filtering process. The question always emerges why do the
carriers not filter this information before they send it on to
the US Government, and many of the carriers I have spoken to are
very happy to do so, but there are a select few carriers, including
British Airways, who are reluctant to filter the information;
they prefer to send the wholesale information to the US Government
and let the US Government take care of the filtering.
Q18 Baroness D'Souza: Why?
Dr Hosein: I believe it is slightly beyond my
competency to say why, but I would guess it is because of the
cost issues.
Q19 Lord Jopling: How on earth do
airlines or anybody else know if you are travelling to New York
or somewhere with your mistress or somebody else? You have got
two separate names and two separate bookings, maybe. How on earth
does anybody know who your travelling companion is?
Dr Hosein: The seating arrangements and whether
at the time of booking you requested to sit with somebody else
would probably be stored within the PNR.
|
