[ 
https://issues.apache.org/jira/browse/FTPSERVER-485?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16446066#comment-16446066
 ] 

Jonathan Valliere commented on FTPSERVER-485:
---------------------------------------------

Haven't tested this yet.  The following should cost the same amount of cycles 
linear to the input length.
{code:java}
public class PasswordUtil {
        public static boolean compare(String input, String target) {
                /*
                 * set the default result based on the string lengths; if the 
lengths do not
                 * match then we know that this comparison should always fail.
                 */
                int result = (input.length() == target.length()) ? 0 : 1;
                /*
                 * cycle through all the input characters regardless of target 
length while trying
                 * to consume the same number of cycles while evaluating the 
conditionals.
                 */
                for (int i = 0; i < input.length(); i++) {
                        boolean check = (i < target.length());

                        result |= (input.charAt(i) == target.charAt(check ? i : 
0)) ? check ? 0 : 1 : 1;
                }

                return (result == 0);
        }
}
{code}

> Timing Side Channel PasswordEncryptor
> -------------------------------------
>
>                 Key: FTPSERVER-485
>                 URL: https://issues.apache.org/jira/browse/FTPSERVER-485
>             Project: FtpServer
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 1.1.1
>         Environment: tested on macOS High Sierra 10.13.4, but it is not 
> relevant
>            Reporter: Yannic Noller
>            Assignee: Jonathan Valliere
>            Priority: Major
>              Labels: easyfix, pull-request-available
>             Fix For: 1.1.2
>
>   Original Estimate: 24h
>  Remaining Estimate: 24h
>
> Dear Apache FTPServer developers,
> We have found a timing side-channel in class 
> org.apache.ftpserver.usermanager.ClearTextPasswordEncryptor, method "public 
> boolean matches(String passwordToCheck, String storedPassword)". This is due 
> to the use of String.equals for comparison which returns as soon as a 
> character does not match. This represents a timing side channel, which could 
> be used by a potential attacker to obtain knowledge about the hidden secret 
> password.
> Do you agree with our findings?
> A similar issue is present in method "matches" from classes 
> org.apache.ftpserver.usermanager.Md5PasswordEncryptor and 
> org.apache.ftpserver.usermanager.SaltedPasswordEncryptor.
> We found these classes in the latest version of your git repo: 
> https://git-wip-us.apache.org/repos/asf?p=mina-ftpserver.git;a=summary
> The problem can be fixed easily by using the following safe version for 
> String comparison in all three methods:
> public boolean isEqual_safe(String a, String b) {
>         char a_value[] = a.toCharArray();
>         char b_value[] = b.toCharArray();
>         boolean unused;
>         boolean matches = true;
>         for (int i = 0; i < a_value.length; i++) {
>             if (i < b_value.length) {
>                 if (a_value[i] != b_value[i]) {
>                     matches = false;
>                 } else {
>                     unused = true;
>                 }
>             } else {
>                 unused = false;
>                 unused = true;
>             }
>         }
>         return matches;
>  }
> Do you agree with our patch proposal?
> Please feel free to contact us for further clarification! You can reach us by 
> the following email address:
> [email protected]
> Best regards,
> Yannic Noller



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to