How to mask Numbers in a variable

[sanjay1]

Hi Team,

I am working on a sas dataset which have account numbers. I want to mask the last 5 digits of the Number with "X";

Here is the dataset;

Data test;

input acct;

cards;

676869845614567

7689645983

;

run;

Desired output

acct

6768698456XXXX

76896XXXXX

Please suggest any approach

Thanks,

Sanjay

[PeterClemmensen]

Do you want to change the actual values or simply apply a format that masks the last 5 digits?

[PeterClemmensen]

If you're looking for a format solution, then do 

proc format;
    picture mask (default=15)
        other = '0000000000XXXXX' (mult=0.00001);
run;

data test;
    value=676869845614567;output;
    value=7689645983; output;
    format value mask.;
run;

Result:

value
6768698456XXXXX
76896XXXXX

[mkeintz]

This is possibly one of the few times it might make sense to use numeric variables instead of character variables for account id's.  @PeterClemmensen 's suggestion allows you to keep actual ACCT values for subsequent matching purposes, while masking a trailing digits when the variable is displayed.  Of course, this approach is really only effective when those data consumers unauthorized to see the actual values don't have direct access to the data variable being formatted.

The trade-off in @Kurt_Bremser 's suggestion is that once the character variable ACCT has been changed (replacing trailing characters with XXXXX), it may no longer be  specific (in fact making it non-specific is the point).  Not only will the modified value no longer be useable as a match key, it could easily create false duplicates within the data set - say if you want the average number of records per account.  So make sure you do all your data aggregation and matching prior to any report generation.

[Kurt_Bremser]

If the goal was just to mask account IDs for output, my method would be sufficient. If the goal is to anonymize the data and keep it usable, then a surrogate key needs to be created.

[mkeintz]

Agreed

[PeterClemmensen]

@mkeintz I agree.

@sanjay1 If value is really character and you do not want to alter the actual value, you can still do this, though it requires a bit more work and the code is not as slick

proc fcmp outlib=work.functions.fun; 
   function mask(string $) $;
      substr(string,max(1, length(string)-4),5)='XXXXX';
      return (string);
   endsub;
run;

options cmplib=work.functions;
proc format;
    value $ mask (default=200) other=[mask()];
run;

data test;
   value='676869845614567'; output;
   value='7689645983'; output;
 
   format value $mask.;
run;
 
proc print data=test;
run;

[ed_sas_member]

Hi @sanjay1 

You can try this 

data want;
	set test;
	acct_new = compress(put(acct,20.));
	substr(acct_new,length(acct_new)-4) = "XXXXX";
run;

(NB: the put() function is not useful if the act variable type is already character.)

[Kurt_Bremser]

Account numbers are not numbers per se, as they are not used for calculations. Therefore store them as character.

Then masking is quite easy:

data have;
input acct :$20.;
cards;
676869845614567
7689645983
;
run;

data want;
set have;
substr(acct,length(acct)-4) = 'XXXXX';
run;

[PGStats]

Just for completeness sake:

data have;
input acct :$20.;
cards;
676869845614567
7689645983
1234
;

data want;
set have;
masked = prxchange("s/\d{5}\s*$/XXXXX/o", 1, acct);
run;

proc print; run;

Obs. 	acct 	masked
1 	676869845614567 	6768698456XXXXX
2 	7689645983 	76896XXXXX
3 	1234 	1234