The Document

6.0 What this chapter gives you#

1. You will be able to state in one sentence what a passport proves and what it does not, and defend that sentence against a fraud investigator.
2. You will be able to name every field in the two machine readable lines at the bottom of a passport, give its character positions, and say which of them the chip repeats word for word.
3. You will be able to compute an ICAO check digit by hand with the 7-3-1 weighting, and say which typing mistakes it always catches and which one in nine of them it misses.
4. You will be able to list the data groups stored on an electronic passport chip, say which two are mandatory, and explain why the file that lists the other files is the one you must never trust.
5. You will be able to explain passive authentication as a chain of signatures, and name the single attack it cannot stop no matter how well it is implemented.
6. You will be able to choose between BAC, PACE, active authentication, chip authentication and terminal authentication, and say for each one what it protects and what it leaves wide open.
7. You will be able to read a European identity card or residence permit and name the regulation that put each element on it, with the dates on which each obligation started and stops.
8. You will be able to look at a security feature under a lamp and say which class of attacker it defeats and which class walks straight past it.
9. You will be able to write down the four separate questions a document check answers, and design a record that stores four answers instead of one flag marked “verified”.

A passport is the most trusted object in ordinary life. People who would not believe a stranger’s word about the weather will believe a small booklet about a stranger’s name, birth date and nationality, and let that booklet decide whether the stranger may board an aircraft, open a bank account or enter a country. It is worth asking what earns that trust, because the honest answer is narrower than almost everybody assumes.

Here is the answer this chapter defends. A passport is not proof of identity. It is proof that an authority once believed something, printed on hard-to-copy paper. Everything in the document, every guilloche line, every check digit, every digital signature, protects the printing and the believing. Nothing in the document protects the truth of what was believed. If a clerk in an office in 2019 was shown a birth certificate that belonged to somebody else and did not notice, then the passport in your hand is completely genuine, perfectly signed, correctly chipped, and wrong. No amount of cryptography applied afterwards can repair that, because the cryptography is a faithful record of a mistake.

That is not a reason to distrust documents. It is a reason to be exact about what a document check tells you, so that you can put the right weight on it and the rest somewhere else. This chapter takes a passport apart: the paper, the two rows of capital letters at the bottom, the chip, the protocols that guard the chip, then identity cards, driving licences and residence permits, then the physical security features and what beats each of them. We finish with the difference between checking a document and checking a person.

Two neighbours to mark and leave alone. How a person’s name gets written into these fields, and why every rule you know about names is false somewhere, is chapter 5. The moment an authority decides that a new person exists and issues a first document, which is where the fraud concentrates, is chapter 7. Here we take that decision as already made, correctly or otherwise, and ask what the resulting object is.

The plain version#

A sealed letter from an office you have never visited#

Imagine a school where children sometimes need to prove who they are to people outside it: a bus conductor, a librarian, a shopkeeper. The head teacher sets up a system. A child comes to the office with their parents. The office looks at whatever the parents bring, writes a short letter, and seals it.

The letter says very little. It says: we, the office of this school, looked at this child on the fourteenth of March, and we believe her name is Anna Eriksson, that she was born on the twenty-fifth of February 2011, and here is a photograph of her face. Then the office presses a metal seal into hot wax, and the seal carries a pattern that only this office owns.

Now Anna carries the letter. When the librarian doubts her, she produces it, and the librarian does three things. She looks at the face in the photograph and looks at Anna. She looks at the wax seal and checks the pattern. She reads the words. If all three go well, Anna gets the book.

A passport is that letter. It is a message from one office to every other office in the world, saying: on this date, we looked at this person and this is what we concluded. The rest of the passport, the whole apparatus of special paper and shifting inks and chips and signatures, exists to stop somebody forging the wax seal or changing the words after sealing. It does nothing about whether the office was right.

Where the letter’s information came from#

Ask the obvious next question. When Anna’s parents came to the office, what did they bring? A different piece of paper, from a different office, saying she was born on that date to those parents. And where did that come from? A hospital told a registrar, and the registrar wrote it in a book.

So the school letter rests on the birth paper, which rests on somebody in a hospital filling in a form. This stack has a name: the papers underneath are called breeder documents, because they breed other documents. Almost every identity document you own sits on top of one piece of paper that nobody alive checked, because the event it records, your birth, happened before you could speak.

This matters more than any cryptography, and here is the reason in one sentence. A birth certificate says a birth happened. It does not say that the person holding it is the person who was born. There is no photograph on it, no fingerprint, no signature from the baby. It is a record of an event, not a description of a body. In England and Wales the point is printed on the certificate itself, in capitals at the bottom, in the words: WARNING: A CERTIFICATE IS NOT EVIDENCE OF IDENTITY.

That gap, between a record of an event and a description of a person, is the softest place in the entire chain, and every serious fraud in this field has gone through it at some point.

The two lines of capital letters#

Turn to the page in a passport with the photograph. At the bottom there are two long rows of capital letters, numbers and angle brackets, in a plain squarish typeface. They look like a mistake. They are the most carefully designed part of the document.

They exist because a border officer with a queue of four hundred people cannot read handwriting, and a machine cannot read a decorative font. So the same facts that appear in the pretty part of the page are repeated at the bottom in a form a cheap scanner can read at speed: fixed positions, one character wide each, a small alphabet, and the angle bracket used as a blank filler so that every field is always exactly the same length.

Here is the second of those two lines, taken exactly as printed in the international specification’s own worked example.

HA672242<6YTO5802254M9601086<<<<<<<<<<<<<<08

Read it in pieces. The first nine characters, HA672242<, are the passport number, padded with a filler because the number is only eight characters long. The next character, 6, is a checking digit, and we will compute it by hand in a moment. Then YTO, a three letter code for the country. Then 580225, which is a date written as year, month, day: 25 February 1958, the holder’s birthday. Then 4, another checking digit. Then M for the sex recorded. Then 960108, the date the passport stopped being valid, 8 January 1996. Then 6, another checking digit. Then fourteen fillers where a national number could have gone, then 0 for that empty field, and finally 8, a checking digit computed over almost the whole line at once.

Nothing here is secret and nothing here is clever. It is simply a way of writing facts so that a machine cannot misread them without noticing.

Doing the checking digit by hand#

The checking digits are the one piece of arithmetic in this chapter that you should do yourself once, because after that you will never again think a check digit is a security feature.

Take a date. The specification’s own example uses 27 July 1952, written as 520727. Write the six digits in a row. Underneath, write the repeating pattern 7, 3, 1, 7, 3, 1. Multiply each pair. Add the six answers. Keep only the last digit of the total.

digits    5    2    0    7    2    7
weights   7    3    1    7    3    1
products 35    6    0   49    6    7

total = 35 + 6 + 0 + 49 + 6 + 7 = 103
last digit of 103 = 3

The check digit is 3, so the date and its check digit are written 5207273. That is the whole algorithm. Letters get numbers too: A counts as 10, B as 11, and so on to Z as 35, and the filler bracket counts as zero. That is all you need to check any of the digits on any passport in the world.

What is it for? If a scanner misreads a 3 as an 8, the total changes, the last digit changes, and the reader knows to ask for another swipe. It is a spelling check, not a lock. Anybody forging a passport simply computes the correct check digits for their invented number, exactly as we just did.

Three things the document does, and one it does not#

Put the plain version together. A modern passport does three jobs well.

It is hard to make. The paper, the inks, the printing presses and the machines that burn the photograph into the page are all expensive, controlled and difficult to imitate, so a casual forger produces something that looks wrong to a trained eye.

It is hard to change. The photograph is fused into the page rather than glued on, the data page is sealed under a film that tears if you lift it, and the two machine readable rows repeat the same facts a second time, so altering a birth year on the pretty half leaves the bottom half disagreeing.

It is hard to fake electronically. Modern passports carry a small chip, and the issuing country signs the contents of that chip with a key nobody else has. A reader can check that signature and know the contents came from that country and have not been edited.

And here is the job it does not do at all: it does not check that the country was told the truth on the day it issued the document. The signature says “this is what we wrote”. It has never said “this is true”.

Where the plain version stops being true#

The seal proves the wax, not the child#

The school letter analogy breaks in the place all such analogies break. The librarian’s three checks, face, seal, words, cover forgery completely and cover fraud not at all.

Suppose a boy takes Anna’s birth paper out of a drawer, walks into the school office and says he is her cousin of the same name. The office follows every rule, checks the paper, photographs the boy, writes the letter and presses the seal. Every subsequent check by every librarian in the country passes, because everything about the letter is real. The seal is genuine, the photograph matches the holder, the words are consistent. Only the belief underneath is false.

The honest version: a document check tells you that a document is genuine and unaltered and, if there is a photograph, that the holder resembles the person it was issued to. It cannot tell you whether the issuing authority was deceived. Documents obtained by deceiving the issuer are called fraudulently obtained genuine documents, and they are invisible to every physical and cryptographic control described in this chapter. They are stopped, if at all, at the issuing counter, which is chapter 7.

The face check is a human comparison, and humans are bad at it#

The plain version said the librarian looks at the photograph and looks at Anna. That comparison is one of the weakest parts of the whole system. Comparing a stranger’s face to a small printed photograph is a task people believe they are good at and are not, and there is now a class of attack designed exactly for it.

In an attack called morphing, two people’s faces are blended into a single image that both of them resemble. One of them, with a clean record, applies for a passport using the blended photograph, and the document is issued to that person entirely genuinely, with a real chip and a real signature. The other person, who could never have obtained a passport, then travels on it, and both the officer and the automated gate accept the face. The attack was demonstrated by Matteo Ferrara, Annalisa Franco and Davide Maltoni of the University of Bologna in a 2014 paper with the deliberately blunt title “The magic passport”. The specification now names morphing as a recognized method of attack.

Nobody may be checking the signature at all#

The plain version said the reader checks the country’s signature. In practice that requires the reader to hold the country’s public key, and holding it means somebody arranged to get it.

ICAO runs a shared directory for exactly this, the Public Key Directory. As of 10 July 2026 it had 111 participants, states and other entities, the most recent being the Maldives. ICAO has 193 Member States. So a large number of passport-issuing countries are not in the shared directory at all, and a receiving country that wants to verify their passports must obtain their keys another way or not verify them.

Worse, there is a structural limit that no amount of participation fixes. Certificates can be revoked; the signed object on an individual chip cannot. If a passport is reported stolen the day after issue, its digital signature stays valid until the certificate that signed it is revoked, and revoking that certificate invalidates every other passport signed with the same key. The specification states the position plainly: only certificates can be revoked, not document security objects. Stolen documents are handled by lost and stolen document databases, not by cryptography.

The chip does not know which book it is glued into#

The plain version implied the chip and the paper are one object. They are not. The chip is a separate component with its own contents and its own signature over those contents. A signature over a set of data proves that the data came from the issuer. It does not prove that this chip is the chip that was originally fitted to this booklet.

This is why the specification separates two ideas that sound identical. Proving the data is authentic is passive authentication. Proving the chip is the original is a different protocol, using a private key held inside the chip that cannot be copied out of it. A country may implement the first and not the second, and many do; where only the first is implemented, a chip’s contents can be copied wholesale onto another chip and will verify perfectly. In 2006 the German researcher Lukas Grunwald demonstrated exactly this.

The two rows of letters are a summary, and summaries lose things#

The machine readable zone is deliberately small: two lines of forty-four characters on a passport. Thirty-nine of those characters carry the entire name. Any name longer than that is cut, by rules written in the specification, and the cut is not marked in any way a machine can rely on.

That has consequences beyond tidiness. A watch list check performed against the machine readable zone is performed against a truncated, transliterated, upper-cased approximation of a name, and matching two such approximations is a probability exercise rather than an equality test. Everything about why that goes wrong belongs to chapter 5, which handles names properly. The point to carry here is that the machine readable zone is not the person’s name. It is a fixed-width encoding of a version of it.

The check digit protects against fingers, not against liars#

We computed one by hand, so this is now easy to state exactly. The 7-3-1 scheme catches every single-character error: change any one character in a field and the check digit always changes, because 7, 3 and 1 all share no factor with 10.

It is not perfect. Swap two neighbouring characters and the check digit changes only sometimes. Working through every possible swap of two adjacent digits in a six-digit date field, fifty of the four hundred and fifty possible swaps leave the check digit unchanged: exactly those where the two swapped digits differ by five. Write 16 as 61 and the check digit does not notice. That is one adjacent transposition in nine missed, which is fine for a scanner and irrelevant to an attacker, who recomputes the digits anyway.

The honest version: check digits in the machine readable zone are error detection for optical character recognition. They are not integrity protection, they are not authentication, and a document whose check digits all pass has told you nothing except that it was probably read correctly.

The technical version#

Breeder documents and the chain that rests on them#

ICAO defines the term in Doc 9303 Part 1: a breeder document is “documentation used as evidence of identity when applying for a travel document”. The definition is procedural rather than physical. A birth certificate is a breeder document when presented in a passport application and an archival record at every other moment.

The chain in most countries runs roughly like this. A birth is notified by a hospital or a family member to a civil registrar, who enters it in a register and issues certified copies on request. Those copies, plus proof of current address and often a countersignature from a professional person, support a first passport or national identity card application. That document then supports a driving licence, a bank account and an employment right check, each of which becomes itself a breeder document for the next.

Civil registration is younger than most people assume, and its age determines how far back the chain can be checked. In England and Wales, the Act for Registering Births Deaths and Marriages in England of 1836 created the General Register Office and a national registration system which took effect on 1 July 1837; registration was made a legal duty only by the Births and Deaths Registration Act 1874, which took effect in 1875. In the United States, the Census Bureau’s birth registration area was established in 1900 and did not cover all states until 1933. In India, the Registration of Births and Deaths Act 1969 created a national scheme, and the Registration of Births and Deaths (Amendment) Act 2023, brought into force on 1 October 2023, made the birth certificate the single document to prove date and place of birth for people born after commencement, for purposes including school admission, electoral rolls and government appointment, and created national and state databases of registered births and deaths.

The specification is candid about the weakness. Doc 9303 Part 2, Appendix D, tells issuing authorities to establish close liaison with the issuers of breeder documents such as birth and marriage certificates and driving licences, and to obtain access to a database of death certificates, so that an application in the name of a dead person can be detected. That last recommendation exists because of a specific, famous fraud: obtaining the birth certificate of a child who died young, whose birth is registered and whose death is registered somewhere else, and building a whole identity on it. It is often named after the 1971 Frederick Forsyth novel that described it. The defence is not a better certificate; it is joining the birth register to the death register, which is an administrative act, not a document feature.

Doc 9303 Part 2 also points issuers at a separate ICAO publication, the TRIP Guide on Evidence of Identity, for establishing the applicant’s entitlement to a document at all. That is the enrolment problem, and it belongs to chapter 7.

The physical passport: dimensions, zones and what is in your hands#

Doc 9303 Part 4 specifies the TD3 form factor, which is the ordinary passport booklet. The data page is 125.00 mm wide by 88.00 mm high, taking the nominal dimensions from ISO/IEC 7810:2019. Total thickness including any laminate must not exceed 0.90 mm, and within the machine readable zone the thickness must not vary by more than 0.10 mm, because a swipe reader drags the page past a fixed optical head. The book must have a cover and a minimum of eight pages, and no page may be added after issuance.

The data page is divided into seven zones. Zones I to VI together form the visual inspection zone, the human readable part; zone VII is the machine readable zone.

+------------------------------------------------+
| Zone I    header: issuing State, "PASSPORT"     |
+----------------+-------------------------------+
|                | Zone II   name, nationality,   |
|  Zone V        |   date of birth, sex,          |
|  portrait      |   place of birth               |
|  min 32 x 26   +-------------------------------+
|  max 45 x 35   | Zone III  document number,     |
|  millimetres   |   date of issue, authority,    |
|                |   date of expiry               |
|                +-------------------------------+
|                | Zone IV   holder's signature   |
+----------------+-------------------------------+
| Zone VII   machine readable zone, 2 lines x 44  |
+------------------------------------------------+
   Zone VI, optional data, sits on the reverse
   of the data page or on an adjacent page.

Zones I, II, III, V and VII are mandatory. The portrait in zone V must be no larger than 45.0 mm by 35.0 mm and no smaller than 32.0 mm by 26.0 mm, and it must be integrated into the page by a secure personalization technology. Stick-on photographs are prohibited outright, and the note that used to permit them was withdrawn from the data element directory in the seventh edition of the specification.

Where a passport carries a contactless chip, the issuer may also print a card access number, a six-digit number used to open a session with the chip. Doc 9303 Part 4 section 4.1.1.2 specifies that it is exactly six numerals with no check digit, because, as the specification puts it, the check is implicitly performed by the protocol.

Two conventions worth separating from the standard. Placing the data page on page 2 or the penultimate page is a recommendation, not a requirement, and placing it on the cover is described as strongly not recommended because covers have historically been attacked by delamination. And the passport number appearing on every sheet is listed as a basic feature in Appendix A to Part 2, which is informative, so it is a very strong convention rather than a rule.

Doc 9303: the twelve parts and how they became one document#

Doc 9303 is in its eighth edition, published in 2021, and is divided into twelve parts. Parts are amended separately: as of August 2026, the consolidated Part 4 carries amendment 1 dated 20 March 2024 and amendment 2 dated 20 February 2026.

The history, taken from the foreword to Part 1, is a useful corrective to the idea that any of this was designed recently. ICAO’s work on machine readable travel documents began in 1968, when the Air Transport Committee of the Council established a Panel on Passport Cards; the panel chose optical character recognition as the reading technology. In 1980 the panel’s work was published as the first edition of Doc 9303, titled A Passport with Machine Readable Capability, and it became the basis for the first machine readable passports, issued by Australia, Canada and the United States. ICAO established the Technical Advisory Group on Machine Readable Travel Documents in 1984. The New Technologies Working Group began work on biometrics in 1998, and that work, largely complete before September 2001, was finalized quickly afterwards and incorporated into volume 2 of the sixth edition of Doc 9303 Part 1 during 2006. In 2005 the then 188 Member States approved a standard requiring all states to issue machine readable passports by 2010 and all non-machine-readable travel documents to have expired by 2015; that standard appears in the thirteenth edition of Annex 9 to the Chicago Convention, published in 2011.

The technical content of Doc 9303 is endorsed by ISO as ISO 7501. The current published part for passports is ISO/IEC 7501-1:2008, the fifth edition, which replaced ISO/IEC 7501-1:2005.

The machine readable zone, field by field#

The TD3 upper line is simple: two characters of document code, three of issuing state or organization, and thirty-nine for the name. The document code begins with P. The second character identifies the type of passport, and this is an area that has just changed. Doc 9303 Part 4 section 4.4 harmonizes the second character: PP for a national or ordinary passport, PD diplomatic, PO official or service, PE emergency, PR refugee, PS stateless, PT alien or non-citizen, PL laissez-passer, PM military. Effective 1 January 2026, any passport issued with a secondary document code must use these values; effective 1 January 2028 every passport must carry one; and passports issued without a harmonized secondary code must expire before 1 January 2038.

Names in the upper line follow rules given in Part 3 section 4.6 and Part 4 section 4.2.3. The primary identifier is separated from the secondary identifier by two filler characters, spaces within a component become a single filler, apostrophes and other punctuation are dropped entirely, and hyphens become a filler. The specification’s own examples, using the fictional issuing state Utopia, are worth reading once:

ERIKSSON, ANNA MARIA
PPUTOERIKSSON<<ANNA<MARIA<<<<<<<<<<<<<<<<<<<

SMITH-JONES, SUSIE MARGARET
PPUTOSMITH<JONES<<SUSIE<MARGARET<<<<<<<<<<<<

O'CONNOR, ENYA SIOBHAN
PPUTOOCONNOR<<ENYA<SIOBHAN<<<<<<<<<<<<<<<<<<

ARKFREITH
PPUTOARKFREITH<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

When a name will not fit in thirty-nine characters it is truncated so that the last character in position 44 is a letter, which is the signal that truncation may have happened. The specification adds a warning which every implementer should read twice: a name that happens to end in a letter at position 44 without being truncated must nevertheless be treated as though it had been truncated, because the two cases are indistinguishable. Names are chapter 5; this is the encoding, not the name.

The lower line is where the identifiers live.

Four details of the lower line catch implementers out. Dates are written YYMMDD with no century, so a reader must apply a windowing rule to decide whether 25 means 1925 or 2025, and the specification gives no rule, which makes it an implementation decision with real consequences for very old and very young travellers. Sex is F, M or the filler character in the machine readable zone, while the visual zone uses X where the issuer does not want to state it. Where the optional data field is unused, position 43 may be either a zero or a filler at the issuer’s option, which means a reader must accept both. And the composite check digit in position 44 is computed over positions 1 to 10, 14 to 20 and 22 to 43, deliberately skipping the nationality code in positions 11 to 13 and the sex in position 21.

The 7-3-1 check digit, worked all the way through#

Doc 9303 Part 3 section 4.9 defines the algorithm: modulus 10 with a continuously repetitive weighting of 731 731, multiply, sum, divide by ten, and the remainder is the check digit. Letters A to Z take the values 10 to 35 and the filler takes zero.

Take the line we met in the plain version, which is the specification’s own composite example from Appendix A to Part 3:

HA672242<6YTO5802254M9601086<<<<<<<<<<<<<<08
123456789 |         |     |       |       |
          10        20    26      34      44

Start with the document number field, positions 1 to 9, HA672242<.

chars     H    A    6    7    2    2    4    2    <
values   17   10    6    7    2    2    4    2    0
weights   7    3    1    7    3    1    7    3    1
products119   30    6   49    6    2   28    6    0

sum = 119+30+6+49+6+2+28+6+0 = 246
246 mod 10 = 6

Position 10 holds 6. It agrees. The same method gives 4 for the date of birth 580225, 6 for the expiry 960108, and 0 for the fourteen fillers in the optional data field.

Now the composite. Concatenate positions 1 to 10, then 14 to 20, then 22 to 43, which gives a string of thirty-nine characters, and restart the weighting from 7 at its first character:

HA672242<6 5802254 9601086<<<<<<<<<<<<<<0

products, first ten:
  119  30   6  49   6   2  28   6   0  42   = 288
products, next ten:
   15   8   0   6   2  35  12   9  42   0   = 129
products, remaining nineteen:
    1   0  24   6   then zeros              =  31

total = 288 + 129 + 31 = 448
448 mod 10 = 8

Position 44 holds 8. The line is internally consistent. Here is the whole algorithm in code, which is short enough that there is no excuse for getting it wrong:

def value(ch):
    if ch == '<':
        return 0
    if ch.isdigit():
        return int(ch)
    return ord(ch) - 55          # 'A' -> 10 ... 'Z' -> 35

def check_digit(field):
    weights = (7, 3, 1)
    total = 0
    for i, ch in enumerate(field):
        total += value(ch) * weights[i % 3]
    return total % 10

line2 = "HA672242<6YTO5802254M9601086<<<<<<<<<<<<<<08"
composite = line2[0:10] + line2[13:20] + line2[21:43]
assert check_digit(line2[0:9])   == int(line2[9])
assert check_digit(line2[13:19]) == int(line2[19])
assert check_digit(line2[21:27]) == int(line2[27])
assert check_digit(composite)    == int(line2[43])

Two honest observations about the scheme’s strength. Because 7, 3 and 1 are each coprime with 10, any single-character substitution always changes the check digit, so single errors are detected with certainty. Adjacent transpositions are not: enumerating all four hundred and fifty adjacent swaps in a six-digit numeric field, fifty of them leave the check digit unchanged, which is one in nine. Those fifty are exactly the swaps of two digits differing by five, because the weight differences 4, 2 and 6 all share a factor with 10. Write 160000 as 610000 and both produce the check digit 5.

None of this is security. Anybody constructing a false machine readable zone runs the same twelve lines of code we just wrote.

The chip: files, data groups and the document security object#

An electronic passport carries a contactless integrated circuit conforming to ISO/IEC 14443, with a minimum storage capacity of 32 kB specified in Doc 9303 Part 10 section 4.5.4. The chip presents an ISO/IEC 7816-4 file system. The passport application is selected by its application identifier, A0 00 00 02 47 10 01. The minimum command set is SELECT and READ BINARY; the security protocols add GET CHALLENGE, EXTERNAL AUTHENTICATE, INTERNAL AUTHENTICATE, MANAGE SECURITY ENVIRONMENT and GENERAL AUTHENTICATE.

The contents are organized as sixteen data groups plus two special files.

Each file has a short elementary file identifier and a template tag: DG1 is short identifier 01 and tag 61, DG2 is 02 and 75, DG3 is 03 and 63, DG15 is 0F and 6F, EF.COM is 1E and 60, and EF.SOD is 1D and 77.

DG1 is the machine readable zone repeated exactly, including filler characters, as a single composite data element in tag 5F1F: two characters of document code, three of issuing state, thirty-nine of name, nine of document number, its check digit, three of nationality, six of birth date, its check digit, one of sex, six of expiry, its check digit, fourteen of optional data, its check digit and the composite check digit. That is why comparing the printed zone with the chip’s DG1 is a meaningful test: they should be identical strings.

DG14 and DG15 are conditional in a precise sense. DG14 is required if the chip supports chip authentication or the generic or integrated mapping of PACE, and it carries SecurityInfos, a set of structures each naming a protocol by object identifier. DG15 is required if the chip supports active authentication, and it carries the public key.

EF.SOD is the object that makes the rest trustworthy. It is a cryptographic message syntax SignedData structure, produced in distinguished encoding rules, whose signed content is an LDSSecurityObject: a version, a hash algorithm identifier, and a sequence of DataGroupHash values, each pairing a data group number with the hash of that data group’s entire file contents. The upper bound ub-DataGroups is 16. The signer’s certificate, the document signer certificate, must be included in the structure.

Now the trap, and it is the single most useful thing in this section. EF.COM lists which data groups are present, and EF.COM is not covered by EF.SOD. It is unsigned. An inspection system that reads EF.COM to decide what to read, and then only verifies what EF.COM told it about, can be led by the nose. This is the basis of the index manipulation attack demonstrated by the Dutch researcher Jeroen van Beek in 2008, which produced chips carrying content that passed inspection software. The specification’s own remedy is blunt: it recommends that inspection systems relying on EF.COM be modified to use the security object instead, as soon as possible. Read the list of hashed data groups from the signed object, not the list of files from the unsigned one.

One live date. ISO/IEC 39794 has succeeded the 2005 first edition of ISO/IEC 19794 for encoding biometrics. Doc 9303 Part 10 sets the transition: inspection systems had to handle ISO/IEC 39794 data by 1 January 2026, issuers may use either format between 2026 and 2030, and from 1 January 2030 issuers must use ISO/IEC 39794. Biometric data in the old encoding sits in data object tag 5F2E; in the new encoding it sits in tag 7F2E. As of August 2026 we are inside the four-year transition window, which means a reader must handle both.

Passive authentication, and the two ways to authenticate the chip#

Passive authentication is the baseline and it is mandatory. Doc 9303 Part 11 section 5.1 gives the inspection procedure. Read EF.SOD, which must contain the document signer certificate. Build and validate a certification path from a trust anchor to that certificate. Use the verified document signer public key to check the signature on EF.SOD. Read the data groups you need. Hash each one and compare with the corresponding hash inside EF.SOD.

The trust anchor is the issuing country’s country signing certification authority, a self-signed root held offline, which issues certificates to one or more document signers, the keys that actually sign passports. The mechanics of certificates and revocation are Volume III of this book; what matters here are the operating parameters, which are unusually aggressive compared with ordinary web certificates.

The logic of the asymmetry is worth stating. A document signer key is used for at most three months, so a compromise exposes at most one quarter of a country’s output; but the certificate must remain valid for about ten years, because passports signed with it stay in circulation for ten years and must still verify. Country signing authorities must issue a certificate revocation list at least every 90 days even if nothing has been revoked, and may issue more often but not more often than every 48 hours. Doc 9303 Part 12 recommends that a new list be issued within 48 hours of a revocation notification, and that receiving states distribute certificates and lists to all border posts within 48 hours.

Passive authentication proves that the contents are authentic and unchanged. It does not prevent an exact copy of the chip’s contents, and it does not prevent chip substitution. Two protocols address that.

Active authentication puts a key pair on the chip. The public key lives in DG15 and is therefore hashed into EF.SOD and signed by the issuer; the private key lives in protected memory and never leaves. The reader sends an eight-byte nonce with INTERNAL AUTHENTICATE, the chip signs a message built from that nonce, using ISO/IEC 9796-2 digital signature scheme 1 for RSA, and the reader verifies. If it verifies, the security object was read from the genuine chip.

Active authentication has a privacy defect the specification documents in its own informative Appendix C, under the name challenge semantics. The chip signs whatever bytes it is given without knowing what they mean. A terminal can therefore construct the challenge as its own signature over an identifier, a date, a time and a location, and the chip’s response then becomes a transferable, cryptographically undeniable proof that this document was at that place at that time. The specification observes that the worst case is active authentication without any access control at all, which would allow a tracking network built from readers in public places whose logs cannot be repudiated.

Chip authentication solves the same problem without that defect. It is an ephemeral-static Diffie-Hellman key agreement: the chip supplies a static public key and domain parameters, the terminal generates an ephemeral pair, both derive a shared secret, and secure messaging restarts with keys derived from it. Because the transcript is not transferable to a third party, there are no challenge semantics. The chip’s static public key is authenticated by passive authentication, so the specification is explicit that passive authentication must be performed together with chip authentication before the chip may be considered genuine. Chip authentication is required for the optional LDS2 applications introduced in the eighth edition.

BAC, PACE and EAC: what each one protects and what it does not#

Adding a radio interface to a passport created two attacks that paper never had: skimming, reading the chip without the holder’s knowledge, and eavesdropping, listening to a legitimate session from a distance. Access control exists for these two, and for nothing else.

Basic access control derives keys from data printed on the page, so that a reader must have seen the page. The inspection system reads a string the specification calls the MRZ information: the document number with its check digit, the date of birth with its check digit and the date of expiry with its check digit. For our worked line that string is:

HA672242<6  +  5802254  +  9601086
= HA672242<658022549601086

The most significant 16 bytes of the SHA-1 hash of that string form the key seed, from which the two document basic access keys are derived. The protocol itself is a three-pass challenge-response following ISO/IEC 11770-2 key establishment mechanism 6, with two-key triple DES in cipher block chaining mode with a zero initialization vector, and a checksum using ISO/IEC 9797-1 MAC algorithm 3 with padding method 2 and an eight-byte tag. Nonces are eight bytes; keying material is sixteen. The session keys come from the exclusive-or of the two sides’ keying material.

The weakness is arithmetic, not implementation, and Doc 9303 Part 11 states it in an informative appendix. For a ten-year document, the maximum strength of the derived keys is about 56 bits where the document number is numeric and about 73 bits where it is alphanumeric, and the higher figure assumes document numbers are random, which they usually are not. In 2005 the researcher Marc Witteman showed that Dutch passport document numbers were predictable, which reduces the real entropy far below the theoretical maximum.

PACE, password authenticated connection establishment, replaces the symmetric derivation with a password-authenticated Diffie-Hellman agreement. The password may be the same MRZ information, or the six-digit card access number, which has the practical advantage that a human can type it. The point of the design is that the strength of the session keys is independent of the entropy of the password: with 256-bit elliptic curves and AES-128, a common choice, the session keys carry 128 bits regardless of whether the password had six digits or sixty. Three mappings exist: generic mapping, integrated mapping, and chip authentication mapping, which folds chip authentication into the same exchange.

The migration is scheduled, and the dates are now close. Doc 9303 Part 11 section 4.1 requires that passports issued from 1 January 2027 implement PACE, either alone or alongside basic access control; that no passport issued from 1 January 2028 support basic access control at all; and that all passports supporting basic access control be out of circulation by 1 January 2038. Chips with no access control at all, permitted by earlier editions, are deprecated in the eighth edition, although inspection systems must still handle them.

Extended access control is a different thing again, and the confusion around it is worth clearing up. It is not an ICAO protocol in origin. It is specified in the German federal office for information security’s technical guideline BSI TR-03110, and Doc 9303 Part 11 reproduces the compatible parts and describes the differences in an informative appendix. Its purpose is to protect the secondary biometrics, fingerprints in DG3 and iris images in DG4, which are far more sensitive than a facial image and which no ordinary reader should be able to obtain. It combines chip authentication with terminal authentication: the terminal presents a chain of card-verifiable certificates starting from a country verifying certification authority whose key the chip already holds, the chip verifies the chain and extracts the terminal’s public key, the chip issues a challenge, and the terminal signs the challenge together with an identifier for the chip and a compressed form of the ephemeral key used to set up the session. The chip then grants exactly the access rights encoded in the terminal’s certificate, expressed as bits in a certificate holder authorization template, with one bit for DG3 and one for DG4.

Read that table as a whole and one line stands out. Nothing on it prevents an attacker who has both the physical document and time from producing a copy of the chip and a copy of the page. Access control means an attacker must see the document; it never meant an attacker who has seen the document is stopped.

Cards: identity cards, driving licences and residence permits#

The card formats reuse everything above with a smaller canvas. A TD1 card is ISO/IEC 7810 ID-1 size, the same as a bank card, and its machine readable zone is three lines of thirty characters, specified in Doc 9303 Part 5. A TD2 document uses two lines of thirty-six, in Part 6. The squeeze shows: on TD1 the name gets thirty characters instead of thirty-nine, and a document number longer than nine characters overflows into the optional data field on the second line, with a filler in the check digit position signalling the overflow.

In the European Union, Regulation (EU) 2019/1157 of 20 June 2019, published in the Official Journal on 12 July 2019 and applying from 2 August 2021, set common requirements for national identity cards. Cards must be ID-1 format with a machine readable zone, must follow Doc 9303 and specifically Part 5 for data elements, must carry the two-letter country code printed in negative in a blue rectangle encircled by twelve yellow stars, and must include a highly secure storage medium holding a facial image and two fingerprints. Children under twelve may be exempted from giving fingerprints and children under six must be. Validity runs from five to ten years, with exceptions for minors and for people aged seventy and over. Biometric data captured for personalization must be erased no later than 90 days from the date of issue.

Two of those dates have now passed, and the August 2026 one passed this month: identity cards that did not meet the minimum security standards in Doc 9303 Part 2, or that lacked a functional machine readable zone, ceased to be valid at their expiry or by 3 August 2026, whichever came first.

There is a legal complication that a careful writer must not skip. On 21 March 2024 the Grand Chamber of the Court of Justice of the European Union, in Case C-61/22, RL v Landeshauptstadt Wiesbaden, declared Regulation 2019/1157 invalid because it had been adopted on the wrong legal basis, under the ordinary legislative procedure rather than the special procedure requiring unanimity in the Council. The Court nonetheless maintained the Regulation’s effects until a replacement adopted on the correct basis enters into force, and in any event no later than 31 December 2026. The Commission brought forward a proposal for a Council Regulation reproducing the text, circulated as Council document 6281/25 in February 2025. [UNVERIFIED: whether the replacement Council Regulation had been adopted and published by August 2026.] The Court separately held that the obligation to store two fingerprints is compatible with the Charter, so the substance survived even though the instrument did not.

Residence documents are the same machinery under a different instrument. Council Regulation (EC) No 1030/2002 of 13 June 2002 laid down a uniform format for residence permits for third-country nationals; Regulation (EU) 2017/1954 amended it, and Commission Implementing Decision C(2018) 7767 carries the technical specifications. Article 7 of Regulation 2019/1157 requires member states to use that same format for residence cards issued to non-EU family members of Union citizens, bearing the title “Residence card” or “Permanent residence card” and one of the standardized codes “Family Member EU Art 10 DIR 2004/38/EC” or “Family Member EU Art 20 DIR 2004/38/EC” in data field 10.

Driving licences are standardized separately, by ISO/IEC 18013. Part 1, in its 2018 edition, covers the physical characteristics and the human readable data; part 2, 2020, covers machine readable technologies; part 3, 2017, covers access control, authentication and integrity validation, deliberately mirroring the passport mechanisms. The mobile driving licence of ISO/IEC 18013-5:2021, which is a different animal entirely, belongs to chapter 53.

The driving licence deserves one paragraph of its own because of what it became. In the United States, which has no national identity card, the state driving licence is the de facto identity document, so identity standards for travel were in effect set by fifty motor vehicle agencies. The REAL ID Act of 2005 imposed federal minimum standards for licences accepted for federal purposes including boarding a commercial aircraft. Enforcement, delayed repeatedly for two decades, began on 7 May 2025, with phased implementation to full enforcement on 5 May 2027. The lesson generalizes: when a document designed for one purpose becomes the identity document, the standards of that original purpose become the identity standards for everybody.

Security features, and what defeats each#

Doc 9303 Part 2 divides features into basic ones considered essential and additional ones from which states are encouraged to choose. It also divides inspection into three levels, and this is the frame that makes the whole subject legible: level 1 is cursory examination at the point of use, using easily identifiable visual or tactile features; level 2 is examination by trained inspectors with simple equipment; level 3 is inspection by forensic specialists.

A feature is only as good as the level of inspection that will actually happen, and a feature detectable only at level 3 does nothing at a hotel reception desk.

A guilloche is defined in Doc 9303 Part 1 as a pattern of continuous fine lines, usually computer generated, forming a unique image that can only be accurately re-originated by access to the equipment, software and parameters used to create the original design. The security is in the re-origination: you cannot derive the generating parameters from the printed result, so a copier reproduces the appearance and not the structure. A two-colour guilloche superimposes two elements in contrasting colours. Where the pattern is computer generated, the specification requires that no pixel structure be detectable in the printed image, which is precisely the artefact a scan-and-reprint attack leaves behind.

Optically variable features change appearance with the viewing angle. The family includes ink with optically variable properties, which shifts colour as you tilt it; diffractive optically variable image devices, which are holograms and their relatives; lenticular features, where a lens structure is built into the surface; and latent images formed in intaglio relief that appear only at a grazing angle. Their common virtue is that they defeat any attack that passes through a flat capture: a photocopier, a scanner, a photograph. Where a travel document is made entirely of plastic, the specification requires optically variable features, precisely because plastic cannot carry many of the paper-based defences.

Laser engraving is defined as a process whereby personalized data are burned into the substrate with a laser. The point is that the data is inside the card body rather than printed on its surface, so altering it means getting inside a fused polycarbonate laminate, which generally destroys the card. Laser perforation, its relative, punches numbers or images through the substrate, which is why the passport number on a modern booklet is often visible as pinholes through several pages and tapers in size from front to back.

Read the right-hand column. Every one of these features is defeated by one of exactly three things: a level of inspection that never happens, possession of genuine blank stock, or a genuine document obtained by deceiving the issuer. That is not a criticism of the features. They do their job, which is making counterfeiting expensive. It is a statement about which risks remain after they have all worked perfectly.

Doc 9303 Part 2 names the categories of attack directly. Counterfeit is the creation of all or part of a document that resembles a genuine one. Fraudulent alteration, also called forgery, is the alteration of a genuine document, with the portrait as the prime target; photo substitution is named as one of the most frequent types of document fraud, which is why stick-in photographs are no longer permitted at all. Whole-page substitution, in which the entire data page is removed and replaced, is treated separately, with countermeasures that are mostly about the sewing: thread that is difficult to remove without evidence, programmable stitching patterns, ultraviolet cured glue on the stitching, serial numbers on every sheet. An impostor is someone representing themselves to be another person. And morphing is listed as its own category.

Why a document check is a document check#

Everything above supports a single conclusion, and it is worth stating as four separate questions rather than one.

Q1  Is this document genuine?
      -> physical features, blank stock control,
         passive authentication of the chip

Q2  Has it been altered since issue?
      -> laminate integrity, MRZ against VIZ,
         DG1 against the printed MRZ, EF.SOD hashes

Q3  Is the person in front of me its holder?
      -> face comparison against DG2 or the printed
         portrait; fingerprint against DG3 if EAC

Q4  Was the issuing authority told the truth?
      -> nothing in the document answers this

Q1, Q2 and Q3 are answerable at the border, at the bank counter or in software. Q4 is not answerable at all by inspecting the document, because a fraudulently obtained genuine document passes Q1, Q2 and Q3 perfectly. The only place Q4 can be attacked is at issuance, which is chapter 7.

This has a direct consequence for how software should record the result. A system that stores a single boolean called identity_verified has thrown away the only information a later investigator will want. Store the four answers, each with what produced it:

{
  "document": {
    "type": "P", "issuer": "UTO",
    "number_hash": "sha256:...",
    "mrz_check_digits": "pass"
  },
  "authenticity": {
    "passive_auth": "pass",
    "trust_anchor": "csca-uto-2021",
    "chip_auth": "not_supported",
    "crl_checked": true, "crl_age_days": 12
  },
  "holder_binding": {
    "face_match": "pass", "score": 0.91,
    "source": "DG2", "morph_check": "not_run"
  },
  "issuance_confidence": "unknown"
}

The last line is the honest one, and almost no production system writes it. It is unknown because the document cannot know it. Making issuance_confidence explicit forces the designer to decide what else in the process is supposed to establish it, rather than letting a green tick on a document scan quietly stand in for an answer nobody obtained.

Two practical rules follow from the same logic, and both are cheap. First, compare the printed machine readable zone with DG1 on the chip, character for character; they are specified to be identical, and a mismatch is a strong signal that chip and book do not belong together. Second, record which checks were skipped, not only which failed. A passive authentication that was not attempted because the country’s trust anchor was missing is a different world from a passive authentication that passed, and a log that stores “pass” for both is worse than no log.

6.98 Common wrong ideas#

Wrong: A passport proves who somebody is. Right: It proves that an issuing authority, on some past date, concluded something about a person and printed the conclusion on a substrate that is hard to copy. Physical features and digital signatures protect the printing and the conclusion in transit; nothing in the document protects the correctness of the conclusion, which is why a fraudulently obtained genuine document passes every check described in Doc 9303.

Wrong: A birth certificate is a strong identity document because it is the original record. Right: It is a record of an event, not a description of a person: no photograph, no biometric, nothing linking the paper to the body presenting it. The certified copy issued in England and Wales says so on its own face, in the words WARNING: A CERTIFICATE IS NOT EVIDENCE OF IDENTITY, and the defence against building an identity on a dead child’s certificate is joining the birth register to the death register, not improving the paper.

Wrong: If the machine readable zone’s check digits all pass, the data is genuine. Right: The 7-3-1 modulus 10 scheme in Doc 9303 Part 3 section 4.9 is error detection for optical character recognition. It catches every single-character misread and misses one adjacent transposition in nine, namely swaps of digits differing by five. A forger computes correct check digits in a few lines of code, so a passing check digit means the line was probably read correctly and nothing more.

Wrong: A chip that verifies cryptographically must be the original chip. Right: Passive authentication proves that the signed data came from the issuer and was not edited; the specification itself states it does not prevent an exact copy of the chip’s contents or chip substitution. Only active authentication, using the private key in DG15, or chip authentication, using the static key pair indicated in DG14, proves the chip itself, and both are optional. Lukas Grunwald demonstrated copying passport chip data onto a standard smart card in 2006.

Wrong: Reading EF.COM tells you which data groups are on the chip. Right: EF.COM is not covered by the document security object and is therefore unsigned, so its list can be edited. Doc 9303 Part 10 recommends that inspection systems relying on EF.COM be changed to use the security object as soon as possible. Take the list of data groups from the signed LDSSecurityObject, and treat any file present on the chip but absent from that list as untrusted.

Wrong: Basic access control encrypts the session, so skimming is solved. Right: The keys are derived by SHA-1 from the document number, date of birth and date of expiry with their check digits, which for a ten-year document gives at most about 56 bits of entropy for a numeric document number and about 73 bits for an alphanumeric one, and less when numbers are sequential, as Marc Witteman showed for Dutch passports in 2005. Doc 9303 Part 11 requires PACE from 1 January 2027, forbids issuing basic access control chips from 1 January 2028, and requires them out of circulation by 1 January 2038.

Wrong: Extended access control is part of the ICAO passport standard. Right: Extended access control originates in the German technical guideline BSI TR-03110 and is a European deployment; Doc 9303 Part 11 specifies compatible chip and terminal authentication protocols and describes the differences from TR-03110 in an informative appendix. It exists to keep fingerprints in DG3 and iris data in DG4 away from readers without a valid card-verifiable certificate, and it protects nothing else.

Wrong: A stolen passport can be revoked like a compromised certificate. Right: Doc 9303 Part 12 states that only certificates can be revoked, not document security objects, so revoking a stolen passport’s signature would mean revoking the document signer certificate and invalidating every other passport signed by that key during its usage period. Stolen documents are handled by databases of lost and stolen travel documents, which is an operational control at the border and not a property of the document.

Wrong: Security features such as holograms and guilloches stop passport fraud. Right: They raise the cost of counterfeiting and of altering a genuine document, and they are effective at that. They do nothing about the three residual routes: an inspection level that never happens, stolen genuine blank stock, and a genuine document obtained by deceiving the issuer. Doc 9303 Part 2’s own three levels of inspection make the first of those explicit.

Wrong: If the face matches the photograph, the holder is the rightful holder. Right: Face comparison by humans against small printed portraits is unreliable, and the morphing attack described by Ferrara, Franco and Maltoni in 2014 produces an image that two different people both plausibly match, so a genuinely issued document can be shared. Doc 9303 Part 2 lists morphing as a recognized attack category, and countermeasures belong at enrolment, in the capture and checking of the submitted image.

6.99 Chapter summary in 20 lines#

1. A passport is evidence that an authority once believed something, printed on a substrate that is hard to copy, and it is nothing more than that.
2. Every identity document rests on breeder documents, and the bottom of that stack is usually a birth certificate that describes an event and not a body.
3. Civil registration is recent: England and Wales from 1 July 1837 and compulsory only from 1875, the United States complete across all states only in 1933.
4. ICAO began work on machine readable travel documents in 1968 and published the first edition of Doc 9303 in 1980, adopted first by Australia, Canada and the United States.
5. Doc 9303 is in its eighth edition of 2021 in twelve parts, amended part by part, with Part 4 amendment 2 dated 20 February 2026.
6. The TD3 data page is 125.00 mm by 88.00 mm in seven zones, with the portrait integrated into the page and stick-on photographs prohibited.
7. The TD3 machine readable zone is two lines of forty-four characters using only A to Z, 0 to 9 and the filler character, printed in OCR-B.
8. The lower line holds document number, nationality, date of birth, sex, date of expiry and optional data, with five check digits in fixed positions.
9. Check digits use modulus 10 with the repeating weights 7, 3, 1, letters valued 10 to 35 and the filler valued zero, and the composite digit skips positions 11 to 13 and 21.
10. The scheme detects every single-character error and misses one adjacent transposition in nine, and it is error detection rather than security.
11. Document codes are being harmonized: from 1 January 2026 any secondary code must follow the table, from 1 January 2028 all passports must carry one, and unharmonized passports must expire before 1 January 2038.
12. The chip holds sixteen data groups, of which DG1, the machine readable zone, and DG2, the facial image, are mandatory, plus EF.COM and the signed EF.SOD.
13. EF.SOD is a signed structure containing one hash per data group, and EF.COM, which lists the files, is unsigned and must never be trusted.
14. Passive authentication verifies a chain from a country signing certification authority through a document signer to EF.SOD, and proves the data but not the chip.
15. Document signer private keys are used for at most three months while their certificates stay valid about ten years, and revocation lists must be issued at least every 90 days.
16. Active authentication proves the chip but creates challenge semantics, a transferable proof of presence; chip authentication proves the chip without that defect.
17. Basic access control derives keys from printed data with about 56 to 73 bits of entropy, and PACE replaces it with a password-authenticated key agreement giving 128-bit session keys.
18. PACE becomes mandatory for passports issued from 1 January 2027, basic access control may not be issued from 1 January 2028, and it must be out of circulation by 1 January 2038.
19. Extended access control from BSI TR-03110 protects only fingerprints and iris images, using terminal authentication with card-verifiable certificates carrying explicit access rights.
20. A document check answers whether the document is genuine, whether it was altered and whether the holder matches, and it cannot answer whether the issuer was told the truth.

Chapter sources: ICAO Doc 9303, Machine Readable Travel Documents, Eighth Edition, 2021, in twelve parts, consulted in the consolidated English editions published by ICAO and read in August 2026. Part 1, Introduction, for the foreword history of the 1968 Panel on Passport Cards, the 1980 first edition titled A Passport with Machine Readable Capability and its adoption by Australia, Canada and the United States, the establishment of the TAG/MRTD in 1984, the New Technologies Working Group from 1998, the 2005 approval by the then 188 Member States of the 2010 and 2015 deadlines published in the thirteenth edition of Annex 9 to the Chicago Convention in 2011, the endorsement of the technical content by ISO as ISO 7501, and section 4.2 for the definitions of breeder document, guilloche design, laser engraving, laser perforation, latent image, microprint, anti-scan pattern, intaglio, tactile feature and the three levels of inspection. Part 2, section 2 for the classification of fraudulent attack as counterfeit, fraudulent alteration, impostor, spoofing and morphing, Appendix A sections A.5.1 to A.5.5.2 for substrate, security printing, ink, numbering, personalization and whole-page substitution, and Appendix D for liaison with breeder document issuers and access to death records. Part 3, section 4.6 for name representation, section 4.8 for the YYMMDD date format, section 4.9 for the modulus 10 check digit with repeating weighting 731, and Appendix A examples 1, 2 and 3, from which the worked line HA672242<6YTO5802254M9601086 with composite check digit 8 and the date example 5207273 are taken verbatim. Part 4, Specifications for Machine Readable Passports and other TD3 Size MRTDs, incorporating amendment 1 of 20 March 2024 and amendment 2 of 20 February 2026, sections 2.2 to 2.6 for dimensions and thickness, section 3.1 for the seven zones, section 4.1.1.1 for the portrait size limits, section 4.1.1.2 for the six-digit card access number, sections 4.2.2.1 and 4.2.2.2 for the upper and lower machine readable line structures, section 4.2.3 for name truncation and its Utopia examples, section 4.2.4 for check digit positions, and section 4.4 for the harmonized document codes with their 2026, 2028 and 2038 dates. Part 10, Logical Data Structure, section 4.1 for the application identifier A0 00 00 02 47 10 01, section 4.5.4 for the 32 kB minimum capacity, section 4.5.6 for the ISO/IEC 39794 transition dates of 1 January 2026 and 1 January 2030 and the 5F2E and 7F2E tags, section 4.6.1 for EF.COM and the recommendation to use the security object instead, section 4.6.2 for EF.SOD and the LDSSecurityObject ASN.1 profile with ub-DataGroups of 16, table 38 for short elementary file identifiers and tags, and section 4.7 for data groups 1 to 16. Part 11, Security Mechanisms for MRTDs, table 1 for the summary of methods and their limits, section 4.1 for compliant configurations and the PACE transition dates of 1 January 2027, 1 January 2028 and 1 January 2038, section 4.3 for basic access control including the SHA-1 key seed, ISO/IEC 11770-2 mechanism 6, two-key triple DES and ISO/IEC 9797-1 MAC algorithm 3, section 4.4 for PACE and its three mappings, section 5.1 for the passive authentication inspection process, section 6.1 for active authentication and ISO/IEC 9796-2 digital signature scheme 1, section 6.2 for chip authentication, section 7.1 for terminal authentication, Appendix A for the 56-bit and 73-bit entropy estimates and the 128-bit PACE session keys, Appendix C for challenge semantics, and Appendix K for European extended access control, EF.CVCA and the DG3 and DG4 access right bits. Part 12, Public Key Infrastructure for MRTDs, section 4.1 table 1 for country signing certification authority private key usage of 3 to 5 years and certificate validity of 13 to 15 years and document signer private key usage of up to 3 months, and section 4.1.5 for the requirement to issue a certificate revocation list at least every 90 days and no more often than every 48 hours and the statement that only certificates can be revoked and not document security objects. ICAO Public Key Directory participant list, read in August 2026, showing 111 participants with the Maldives joining on 10 July 2026 and the first six on 19 March 2007; ICAO Member States page giving 193 Member States. ISO/IEC 7501-1:2008, fifth edition of 15 April 2008, replacing ISO/IEC 7501-1:2005; ISO/IEC 7810:2019; ISO/IEC 18013-1:2018, 18013-2:2020 and 18013-3:2017 for the ISO-compliant driving licence, and ISO/IEC 18013-5:2021 for the mobile driving licence covered in chapter 53. BSI Technical Guideline TR-03110 for extended access control. Regulation (EU) 2019/1157 of the European Parliament and of the Council of 20 June 2019, Official Journal L 188 of 12 July 2019 page 67, articles 3, 4, 5, 7, 8, 10 and 16, applying from 2 August 2021, with the phase-out dates of 3 August 2023, 3 August 2026 and 3 August 2031; Council Regulation (EC) No 1030/2002 of 13 June 2002 as amended by Regulation (EU) 2017/1954, and Commission Implementing Decision C(2018) 7767. Judgment of the Court of Justice of the European Union, Grand Chamber, of 21 March 2024 in Case C-61/22, RL v Landeshauptstadt Wiesbaden, declaring Regulation 2019/1157 invalid for incorrect legal basis while maintaining its effects until a replacement enters into force and at the latest until 31 December 2026, and the Commission proposal circulated as Council document 6281/25 of February 2025. Council Regulation (EC) No 2252/2004 of 13 December 2004, applied as regards the facial image by 28 August 2006 and fingerprints by 28 June 2009. REAL ID Act of 2005, with enforcement begun by the Transportation Security Administration on 7 May 2025 and phased to full enforcement on 5 May 2027. The Act for Registering Births Deaths and Marriages in England 1836 and the Births and Deaths Registration Act 1874; the Registration of Births and Deaths Act 1969 of India and the Registration of Births and Deaths (Amendment) Act 2023 in force from 1 October 2023, as summarized by PRS Legislative Research. Matteo Ferrara, Annalisa Franco and Davide Maltoni, “The magic passport”, IEEE International Joint Conference on Biometrics, 2014. Published accounts of Lukas Grunwald’s 2006 demonstration of copying passport chip data to a standard ISO/IEC 14443 card, of Marc Witteman’s 2005 finding that Dutch document numbers were predictable, and of Jeroen van Beek’s 2008 index manipulation demonstration.