> On June 21, 2018, 7:44 p.m., Velmurugan Periasamy wrote:
> > ugsync/src/main/java/org/apache/ranger/unixusersync/process/UnixUserGroupBuilder.java
> > Lines 543 (patched)
> > <https://reviews.apache.org/r/67694/diff/1/?file=2044228#file2044228line543>
> >
> >     MD5 is not recommended anymore.
> 
> Allen Wittenauer wrote:
>     Agree that MD5 shouldn't be used for security purposes, but that isn't 
> the use case here.  Instead, it is only used to generate a simple checksum.  
> Using a more complex (and therefore more CPU intensive) hashing function 
> doesn't have much value.  If someone were to replace /etc/passwd with a file 
> that had an MD5 collision (the reason why MD5 shouldn't be used in the 
> majority of use cases) it would defeat the purpose; this code is only 
> triggered when the MD5s do not match.
> 
> Velmurugan Periasamy wrote:
>     It is a good idea to use something like sha256Hex, so that source code 
> analysis tools such as coverity/fortify do not flag md5Hex usage as 
> vulnerable.
> 
> Cetin Sahin wrote:
>     Dear Velmurugan and Allen,
>     
>     Thanks for your comments. I also do not think using MD5 impose any 
> security risk in this context, but I will be happy to replace the digestion 
> algorithm with a more stronger one if you think it is really necessary. At 
> first place, I was more concerned about the performance rather than the 
> security of /etc/passwd or /etc/group files. If UnixUserGroupBuilder checks 
> the update too frequently (which is configurable in the Ranger context), 
> using more secure digestion algoritm like SHA256 will be computationally 
> heavier without any additional functional benefits.

Thanks Cetin. Sorry for the delayed response. I agree it would be 
computationally heavier, my suggestion is to go with that anyway since 
Coverity/Fortify scans might flag any MD5 usage as potential issue.


- Velmurugan


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/67694/#review205204
-----------------------------------------------------------


On June 21, 2018, 7:16 p.m., Cetin Sahin wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/67694/
> -----------------------------------------------------------
> 
> (Updated June 21, 2018, 7:16 p.m.)
> 
> 
> Review request for ranger.
> 
> 
> Bugs: RANGER-2139
>     https://issues.apache.org/jira/browse/RANGER-2139
> 
> 
> Repository: ranger
> 
> 
> Description
> -------
> 
> Fixed the update detection issue on consecutive updates in 
> UnixUserGroupBuilder. The update detection logic is improved by verifying the 
> checksums in addition to last modification time.
> 
> 
> Diffs
> -----
> 
>   
> ugsync/src/main/java/org/apache/ranger/unixusersync/process/UnixUserGroupBuilder.java
>  ddab6294a 
> 
> 
> Diff: https://reviews.apache.org/r/67694/diff/1/
> 
> 
> Testing
> -------
> 
> 1. Applied the patch to the master branch and verified that all unit tests 
> passed successfully.
> 
> 
> Thanks,
> 
> Cetin Sahin
> 
>

Reply via email to