Computer Science 9608 (Notes)

Chapter: 1.6 Security, privacy and data

Topic: 1.6.2 Data integrity

Data Integrity defines a quality of data, which guarantees the data is complete and has a whole structure.
Data integrity is most often talked about with regard to data residing in databases, and referred to as
database integrity as well. Data integrity is preserved only if and when the data is satisfying all the
business rules and other important rules. These rules might be how each piece of data is related to each
other, validity of dates, lineage, etc. According to data architecture principles, functions such as data
transformation, data storage, meta-data storage and lineage storage must guarantee the integrity of data.
That means, data integrity should be maintained during transfer, storage and retrieval.

If data integrity is preserved, the data can be considered consistent and can be given the assurance to be
certified and reconciled. In terms of data integrity in databases (database integrity), in order to guarantee
that integrity is preserved, you have to ensure that the data becomes an accurate reflection of the
universe it is modeled after. In other words, it must make sure that the data stored in the database
corresponds exactly to the real world details it is modeled after.

Validation and Verification

When data is input to a computer system it is only valuable data if it is correct. If the data is in error in any
way then no amount of care in the programming will make up for the erroneous data and the results
produced can be expected to be unreliable. There are three types of error that can occur with the data on
entry. The first is that the data, while reasonable, is wrong. If your birthday is written down on a data
U U

capture form as 18th of November 1983, it will (except in very rare cases) be wrong. It can be typed into the
P P

computer with the utmost care as “181183”, it can be checked by the computer to make sure that is a
sensible date, and will then be accepted as your date of birth despite the fact that it is wrong. There is no
reason for the computer to imagine that it may be wrong, quite simply when you filled out the original form
you made a mistake. The second type of error is when the operator typing in the data hits the wrong key
U

and types in “181193”, or the equivalent. In this case an error has been made that should be able to be
U

spotted if a suitable check is made on the input. This type of data checking is called a verification
check. The third type of error is when something is typed in which simply is not sensible. If the computer
U U

knows that there are only 12 months in a year then it will know that “181383” must be wrong because it is
not sensible to be born in the thirteenth month. Checks on the sensibility of the data are called validation
checks.

Faulty data:
U

There is very little that can be done about faulty data except to let the owner of the data check it visually
on a regular basis. The personal information kept on the school administration system about you and your
family may well be printed off at regular intervals so that your parents can check to ensure that the stored
information is still correct.

Page 1 of 4
Computer Science 9608 (Notes)
Chapter: 1.6 Security, privacy and data
integrity

Topic: 1.6.2 Data integrity

Verification means tallying the input data with the original data to make sure that there have been no
transcription errors. The standard way to do this is to input the data twice to the computer system. The
computer then checks the two sets of data (which should be the same) and if there is a difference
between the two sets of data the computer knows that one of the inputs is wrong. It won’t know which
one is wrong but it can now ask the operator to check that particular input.

Validation:
U

The first thing is to clear out a common misinterpretation of validation. Specifically, the use of parity bits
to check data. This is NOT validation. Parity bits and echoing back are techniques that are used to check
U U

that data has been transmitted properly within a computer system (e.g. from the disk drive to the
processor), validation checks are used to check the input of data to the system in the first place.

Validation is a check on DATA INPUT to the system by comparing the data input with a set of rules that
the computer has been told the data must follow. If the data does not match up with the rules then there
must be an error. There are many different types of validation checks that can be used to check input in
different applications:

1. Range check. A mathematics exam is out of 100. A simple validation rule that the computer can apply
to any data that is input is that the mark must be between 0 and 100 inclusive. Consequently, a mark of
101 would be rejected by this check as being outside the acceptable range.

2. Character check. A person’s name will consist of letters of the alphabet and sometimes a hyphen or
apostrophe. This rule can be applied to input of a person’s name so that “dav2d” will immediately be
rejected as unacceptable.

3. Format check. A particular application is set up to accept a national insurance number. Each person
has a unique national insurance number, but they all have the same format of characters, 2 letters
followed by 6 digits followed by a single letter. If the computer knows this rule then it knows what the
format of a NI number is and would reject “ABC12345Z” because it is in the wrong format, it breaks the
rule.

4. Length check. A NI number has 9 characters, if more or fewer than 9 characters are keyed in then the
data cannot be accurate.

5. Existence check. A bar code is read at a supermarket check-out till. The code is sent to the main
computer which will search for that code on the stock file. As the stock file contains details of all items
held in stock, if it is not there then the item cannot exist, which it obviously does, therefore the code must
have been wrongly read.

6. Check digit. When the code is read on the item at the supermarket, it consists of numbers. One number
is special; it is called the check digit. If the other numbers have some arithmetic done to them using a
Page 2 of 4
Computer Science 9608 (Notes)
Chapter: 1.6 Security, privacy and data
integrity

Topic: 1.6.2 Data integrity

Error checking and correcting

When you send data across the internet or even from your USB to a computer, you are sending millions
upon millions of ones and zeros. What would happen if one of them got corrupted? Think of this situation:
You are buying a new game from an online retailer and put £40 into the payment box. You click on send
and the number 40 is sent to your bank stored in a byte: 00101000. Now imagine if the second most
significant bit got corrupted on its way to the bank, and the bank received the following: 01101000. You'd
be paying £104 for that game! Error Checking and Correction stops things like this happening. There are
many ways to detect and correct corrupted data, we are going to learn two.

a) Parity: All data is transmitted as bits (0s and 1s). The Number of 1s in a byte must always be either an
odd number or an even number. If two devices that are communicating decide that there will always be an
odd number of 1s, then if a byte is received that has an even number of 1s, an error must have occurred.
E.g. the byte 01011000 has 3 ones in it. 3 is an odd number, so it fits the rule that it must have an odd
number of ones. When it is sent there is an error in transmission so that the first bit is received as a one.
So, the byte received is 11011000. This has 4 ones in it, which is an even number, so there must be an
error. The receiving device would ask for it to be sent again.

Notes:

If two mistakes are made in the same byte they will cancel each other out and the faulty data will be
accepted. This problem can be overcome, and in the same way, a clever way of correcting error
mistakes can be implemented. This method is not part of this course.
Earlier in this course it was said that a byte was the number of bits necessary to hold a character
code. Specifically, an ASCII character uses 8 bits in a byte, giving 256 different characters. This is not
true because one of the bits has to be reserved for a parity bit, the bit that can change to make sure
that the number of ones is always odd. This means that there are 128 different characters possible.
The implication in all the above work is that odd parity is always used. Even parity can equally well be
U U

used, whichever has been agreed between the two devices.

Page 3 of 4
Computer Science 9608 (Notes)
Chapter: 1.6 Security, privacy and data
integrity

Topic: 1.6.2 Data integrity

It is an error-checking technique involving the comparison of a transmitted block check character with
one calculated by the receiving device. Parallel parity is based on regular parity. Parallel parity can detect
whether or not there was an error, furthermore it can detect which bit has flipped. This method is
implemented on a block of data which is made of sum words; the parity bit is then added to the columns
and rows.

Here’s an example:

1 1 1 0 1 0 1 1
1 1 0 0 0 1 1 0
0 0 1 0 0 1 1 1
0 0 0 0 1 0 1 0

b) Check Sum: Data will normally be sent from one place to another as a block of bytes rather than as
individual bytes. The computer can add numbers together without any trouble, so another checking
procedure is to add all the bytes together that are being sent in the block of data. The carry, out of the
byte, is not taken into account, so the answer is an 8 bit number, just like the bytes. This answer is
calculated before the data is sent, and then calculated again when it is received, and if there are no errors
in the transmission, the two answers will match.

Page 4 of 4