Revision as of 14:34, 11 November 2020

General Observations

QR code, content has variable length at around 330 bytes.
Contains readable strings as well as binary parts.
Seems to be a sequence of variable length records, rather than a fixed binary layout, could be some form of TLV encoding (there are some similarities to ASN.1 BER/DER for example).
Records consist of a 1 byte type field, a length field and N content bytes.
The type field does not seem to describe semantics, as values repeat for different values (rather than a data type of some form?). This would suggest that the record order defines their semantics. The presence of null records suggests that too, as well as the same record sequence in all samples.
"European Union Agency For Railways - Technical Document - Digital Security Elements For Rail Passenger Ticketing - TAP TSI TD B.12 - §11 FCB - Flexible Content Barcode" describes something that seems similar, using ASN.1 UPER encoding.

Record Structure

1 byte type, although no idea what that indicates. Largely doesn't match the BER/DER type byte.
Length:
- for values <= 127 byte this is just one byte
- for value larger than 127 this is encoded in two bytes. This is quite different from BER/DER multi-byte length encoding however. It looks like a little endian layout , with the most significant bit of the first length byte being removed and the second byte being shifted by one bit to fill that space.
N bytes of content, with varying data types.

Data Types

Strings

Strings seem to be UTF-8 encoded.
The length is the amount of bytes needed to represent the UTF-8 string, not the amount of characters.

Date/Time

This is largely speculation at this point!

There's 5 7 byte sequences included that could be date/time values. These could be date of purchase/issue, begin of validity, end of validity, traveler birth date. All of those are printed on the ticket.

the first byte is 0x08 in all samples
the first half of the second byte seems to be 0x8 for date fields and 0xC for date/time fields. This would seem consistent with optional ASN.1 UPER field encoding.
there's a surprising amount of entropy in those values, esp. when looking at differences at close-by dates.
differential view shows no obvious correlation beyond multiple tickets, but close-by values do "look" similar nevertheless.
the suspected birthday field is the same in all samples for the same traveler
this does not seem to use a UNIX timestamp
this does not seem to use BCD encoding
this does not seem to use sub-byte per-component encoding

Record Sequence

Nesting Depth	Type Id	Content Type	Meaning	Notes
0	0x0A	date/time, nested record	?	speculative, given the 7 byte field there matches suspected date/time values below
1	0x12	nested record
2	0x0A	2bytes followed by nested records	?	0x08 04
3	0x12	string	ticket type?	"Point-to-point Ticket", "Supersaver Ticket"
2	0x12	string	Departure station
2	0x1A	string	Arrival station
2	0x20	2 byte	?	0x2801 in all samples
2	0x32	string	Via
2	0x38	1 byte	?	0x42 in all samples
2	0x42	7 byte	?	date/time?
2	0x4A	7 byte	?	date/time?
2	0x60	variable, null terminated	?	2-3 bytes in all samples, breaks TLV structure
2	0x7A	string	ticket type or tariff	parenthesis enclosed abbreviations, string is also printed on the ticket
2	0x88	1 byte	?	null, optional
2	0x98	1 byte	?	0x02
2	0xA0	1 byte	?	null
1	0x1A	nested record	traveler information	could also be loyalty program info?
2	0x0A	string	SBB customer id	only the first 8 digits (of 10) though
2	0x12	string	?	128 bit as a 36 byte hex string with for separator dashes (uuid-like formatting)
2	0x1A	string	family name
2	0x22	string	given name
2	0x2A	7 byte	?	date/time? - possibly traveler birth date
2	0x3A	string	tariff information?	"HALBTAX"
1	0x2A	nested record	?
2	0x0A	7 byte	?	date/time?
2	0x10	4 byte	?	followed by 0x200B outside of TLV structures?
1	0x32	nested record	price information?
2	0x0A	string	payment method?	"PCD", "MC"
2	0x12	string	currency	"CHF"
2	0x1A	string	ticket price
1	0x3A	null	?
1	0x42	nested record	train information	not present in unbound tickets
2	0x5A	string	train number
2	0x70	null
2	0x78	null
1	0x48	5byte, no length byte!?!	?	breaks the overall TLV structure?? almost fixed value in all samples
0	0x22	nested record	?
1	0x0A	string	?	4 digit number (3342 in all samples)
1	0x12	string	?	5 digit number (00001 in all samples)
1	0x2A	nested record
2	0x30	nested record		see KDE_PIM/KItinerary/Thalys_Barcode, same structure there
3	0x02	20-21 byte	?	signature?
3	0x02	20-21 byte	?

@@ Line 29: / Line 29: @@
 There's 5 7 byte sequences included that could be date/time values. These could be date of purchase/issue, begin of validity, end of validity, traveler birth date. All of those are printed on the ticket.
 * the first byte is 0x08 in all samples
+* the first half of the second byte seems to be 0x8 for date fields and 0xC for date/time fields. This would seem consistent with optional ASN.1 UPER field encoding.
 * there's a surprising amount of entropy in those values, esp. when looking at differences at close-by dates.
 * differential view shows no obvious correlation beyond multiple tickets, but close-by values do "look" similar nevertheless.