Zhanga Redux

The chronicles of the work and personal life of a boring software developer with an awesome dog.

Darktable + Nvidia OpenCL on Fedora

Thursday, March 20, 2014


Darktable's OpenCL isn't activated by default on Fedora when using the RPMFusion Nvidia drivers, apparently because Darktable can't find the library:

$ darktable -d opencl
(...)
[opencl_init] trying to load opencl library: ''
[opencl_init] could not find opencl runtime library 'libOpenCL'
[opencl_init] no working opencl library found. Continue with opencl disabled
[opencl_init] FINALLY: opencl is NOT AVAILABLE on this system.
[opencl_init] initial status of opencl enabled flag is OFF.

A quick, if hacky, solution is to just make a symlink to Darktable's directory so it can find the library:

# ln -s /usr/lib64/nvidia-304xx/libOpenCL.so.1 /usr/lib64/darktable/libOpenCL.so

(Obviously, adjust the paths based on your system.)

Tags: fedora, darktable | Posted at 01:35 | Comments (0)

Prosper's insecure bank account management feature

Wednesday, December 4, 2013


Prosper Marketplace is one of the two big peer-to-peer lending platforms (the other being Lending Club), where individuals can loan or borrow money from each other. Like any online investment system, it provides a way for users to move money in and out via ACH transfers to regular bank accounts. And to do that, one needs to link one's Prosper account and bank account.

Since we're talking about a financial institution making important account changes, surely they have a secure way to do this, right? Wrong.

Back in March 2012 when I started making loans on Prosper, the process was pretty standard: enter the bank account number into Prosper.com (over HTTPS), wait a few days for some verification deposits of less than $1 each, and then log back into Prosper.com and type in the amounts to verify that you really own the bank account. This seems to work for most financial institutions, but apparently not Prosper, because, in a unique move that harms both security and usability, they've changed it. Now, when trying to add or remove a bank account, Prosper shows the following message:

For security reasons, to add or change your bank account, you must print a fax cover sheet and attach a copy of a cancelled check from the bank account you wish to add.

Click here to print a specially coded fax cover sheet which will speed processing. Specific instructions are printed on the cover sheet.

The instructions say to fax or email a copy of the account holder's driver's license and either a voided check or recent bank statement. And they won't accept this form snail-mailed or hand-delivered (I called to check). This is horrible!

Fax and email are both completely insecure methods of communication. Anyone between me and Prosper can read, copy, and store my driver's license, bank account number, address, and whatever other personal information I might be sending (like any recent transactions shown on the bank statement). Email is probably worse because typically emails are routed through a couple of servers, creating multiple copies along the way.

It's especially scary that Prosper's message begins with "for security reasons" — is their intent to eliminate security? Because that's the only effect on security that I can see. If there is any security benefit to this approach at all, I'd love to hear. Anyway, how can Prosper even verify whether a copy of a driver's license, check, or bank statement is legitimate? Copies of all of these items are trivially forgeable.

Worse, Prosper is requiring its lenders to send private documents over an insecure channel. One example attack, just for illustration, might be the following:

  1. Intercept a legitimate fax/email containing relevant documents. (Example methods: hack an insecure mail server along the route, hack a router, DNS attacks, routing attacks, bribe a Gmail/Hotmail/Yahoo employee, wiretap ...)
  2. Slightly modify the voided check or bank statement to show the same name, but an account number owned by the attacker.
  3. Send Prosper a new email containing the real driver's license with the modified check or bank statement.
  4. The victim probably doesn't notice since the "Bank Accounts" section of the Prosper site only shows the name of the bank and the last few digits of the account number; only the account number will have changed.
  5. Wait for the real user to withdraw some cash out of Prosper.
  6. Profit!

I am extremely disappointed that Prosper has taken such a giant leap backwards in terms of security and am strongly considering withdrawing my funds from the platform and leaving — if this is how their public-facing side operates, then what security nightmares are hidden from view?

Tags: security, prosper | Posted at 23:42 | Comments (0)

$_SERVER['PHP_SELF'] and cross-site scripting

Monday, May 20, 2013


It's tempting to assume that PHP's $_SERVER array mostly contains fields out of the reach of an attacker, since these are "server" variables. However, that's not always the case; in particular, the seemingly innocuous PHP_SELF field can be a vector for cross-site scripting.

For example, consider the following foo.php:

<form method="POST" action="<?php echo $_SERVER['PHP_SELF'] ?>">
    <!-- ...form elements... -->
</form>

If I visit http://www.example.com/foo.php, $_SERVER['PHP_SELF'] will be /foo.php and everything will work correctly.

But what if I visit http://www.example.com/foo.php/"><script>alert('hello');</script> instead? Then the rendered HTML will be:

<form method="POST" action="/foo.php/"><script>alert('hello');</script>">
    <!-- ...form elements... -->
</form>

This allows injection of arbitrary script running under the host site's context, also known as XSS. Two ways to fix this are:

  • Use $_SERVER['SCRIPT_NAME'] instead of $_SERVER['PHP_SELF']. The former is the name of the actual script file and can't normally be manipulated by an attacker.
  • Use htmlspecialchars(), which by default will escape double-quotes and prevent a user-supplied string from breaking out of an HTML attribute context.

By the way, this was pretty surprising behavior to me for two reasons:

  • The documentation of PHP_SELF is misleading: The first sentence says:

    The filename of the currently executing script, relative to the document root.

    It seems odd that PHP would refer to something like /foo.php/"><script>alert('hello');</script> as a "filename."
  • It's pretty bizarre default behavior that PHP will execute /foo.php for a request of /foo.php/bar/baz.

Tags: security, xss, php | Posted at 11:13 | Comments (0)

Displaying Chinese UTF-8 characters in gvim on Windows

Tuesday, April 2, 2013


By default, gvim on my Windows machines just displays question marks, boxes, or garbled characters when I try to open files with Chinese text. The fix was rather simple:

  1. From the gettext project on SourceForge, get libiconv-1.9.1.bin.woe32.zip, which contains bin/iconv.dll. Put that file into gvim's installation directory (for me, it's C:\Program Files (x86)\Vim\vim73).
  2. Put this into vimrc:
    set encoding=utf8
    set guifontwide=NSimSun

Note: I'm using gvim 7.3.46 from the official site.

Credit goes to user Tobbe for this answer on superuser.com, which pointed out the key ingredient (iconv.dll).

Tags: chinese, unicode, vim | Posted at 11:31 | Comments (2)

Decertifying secured Stanford transcript PDF

Friday, March 29, 2013


Stanford (and probably many other schools) provide official transcripts in the form of a signed PDF. This transcript can only be opened using Adobe Acrobat; other readers simply choke on the document and can't open it. This is caused by the use of a proprietary "secured" document signing feature, apparently exclusive to Adobe Reader, which verifies the authenticity of the transcript. (For comparison, Duke University provides regular PDFs that can be opened by any reader, but can be verified by uploading the file to a trusted authority who confirms the document's legitimacy.)

Stanford's PDFs are particularly annoying because you can't take a screenshot with Print Screen (though the Snipping Tool works), Acrobat won't remove the signature, and attempting to use Acrobat Reader print the PDF to CutePDF simply results in a mostly-blank page with the following text:

ERROR: undefined
OFFENDING COMMAND: get
STACK:
/quit
-dictionary-
-mark

Here's how to un-certify the PDF:

  1. Open the secured PDF in Acrobat Reader and "Print to File" (File/Print, click Advanced, and check the box).

  2. Note: this step may violate certain laws. Proceed at your own risk. A comment in the file declares that "Removing the following eleven lines is illegal, subject to the Digital Copyright Act of 1998."

    Open the resulting .ps file in a text editor. Find and delete the following block:

    mark currentfile eexec
    ... stuff here ...
    cleartomark

    Then, use ps2pdf or similar to convert the PostScript file to PDF.

    This entire step can be accomplished with the following:

    $ sed '/mark currentfile eexec/,/cleartomark/d' secured.ps \
          | ps2pdf - unsecured.pdf
  3. For Stanford transcripts, the pages need to be rotated clockwise, so use pdftk:

    $ pdftk input.pdf cat 1-endE output output.pdf

Tags: stanford, pdf, secured | Posted at 14:51 | Comments (2)