[HN Gopher] LaTeXpOsEd: A Systematic Analysis of Information Lea...
       ___________________________________________________________________
        
       LaTeXpOsEd: A Systematic Analysis of Information Leakage in
       Preprint Archives
        
       Author : oldfuture
       Score  : 65 points
       Date   : 2025-10-13 08:33 UTC (14 hours ago)
        
 (HTM) web link (arxiv.org)
 (TXT) w3m dump (arxiv.org)
        
       | SiempreViernes wrote:
       | As far as I can tell they trawled a big archive for sensitive
       | information, (unsurprisingly) found some, and then didn't try to
       | contact anyone affected before telling the world "hey, there are
       | login credentials to be found in here".
        
         | crote wrote:
         | Don't forget giving it a fancy name in the hope that it'll go
         | viral!
         | 
         | I am getting _so_ tired of _every_ vulnerability getting a
         | cutesy pet name trying to pretend being the new Heartbleed  /
         | Spectre / Meltdown...
        
           | wongarsu wrote:
           | Beats having to remember and communicate CVE numbers
        
         | KeplerBoy wrote:
         | It's not like every datapoint comes with the email of the
         | corresponding author.
        
       | mseri wrote:
       | Google has a great aid to reduce the attack surface:
       | https://github.com/google-research/arxiv-latex-cleaner
        
         | Y_Y wrote:
         | I use this before submission and recommend others do too. If ai
         | was in charge of arXiv Id have it integrated as an optional
         | part of the submission process.
        
         | Jaxan wrote:
         | I do this by hand. In the whole process, cleaning up the TeX
         | before submission is a small step. And I like to keep some
         | comments, like explaining how some Tikz figures are made. Might
         | help someone some day.
        
         | andrepd wrote:
         | Yes but then we wouldn't have https://xcancel.com/LeaksPh
        
       | barthelomew wrote:
       | Paper LaTeX files often contain surprising details. When a paper
       | lacks code, looking at latex source has become a part of my
       | reproduction workflow. The comments often reveal non-trivial
       | insights. Often, they reveal a simpler version of the methodology
       | section (which for poor "novelty" purposes is purposely obscured
       | via mathematical jargon).
        
         | seg_lol wrote:
         | Reading the LaTex equations also makes for easier (llm)
         | translation into code rather than trying to read the pdf.
        
       | kmm wrote:
       | I sort of understand the reasoning on why Arxiv prefers tex to
       | pdf[1], even though I feel it's a bit much to make it mandatory
       | to submit the original tex file if they detect a submitted pdf
       | was produced from one. But I've never understood what the added
       | value is in hosting the source publicly.
       | 
       | Though I have to admit, when I was still in academia, whenever I
       | saw a beautiful figure or formatting in a preprint, I'd often try
       | to take some inspiration from the source for my own work,
       | occasionally learning a new neat trick or package.
       | 
       | 1: https://info.arxiv.org/help/faq/whytex.html
        
         | irowe wrote:
         | A huge value in having authors upload the original source, is
         | it divorces the content from the presentation (mostly). That
         | the original sources were available was sufficient for a large
         | majority of the corpus to be automatically rendered into HTML
         | for easier reading on many devices:
         | https://info.arxiv.org/about/accessible_HTML.html. I don't
         | think it would have been as simple if they had to convert PDFs.
        
       | cozzyd wrote:
       | This is why my forarxiv.tex make targets always include a call to
       | latexpand --empty-comments
       | 
       | Though I doubt all my collaborators do something similar.
        
       | sneela wrote:
       | I agree with other comments that this research treads a fine,
       | unethical line. Did the authors responsibly disclose this, as is
       | often done in the security research community? I cannot find any
       | mention of it in the paper. The researchers seem to be involved
       | in security-related research (first author is doing a PhD, last
       | author holds a PhD).
       | 
       | At least arxiv could have run the cleaner [1] before the print of
       | this pre-print (lol). If there was no disclosure, then I think
       | this pre-print becomes unethical to put up.
       | 
       | > leading to the identification of nearly 1,200 images containing
       | sensitive metadata. The types of data represented vary
       | significantly. While device information (e.g., the camera used)
       | or software details (such as the exact version of Photoshop) may
       | already raise concerns, in over 600 cases the metadata contained
       | GPS coordinates, potentially revealing the precise location where
       | a photo was taken. In some instances, this could expose a
       | researcher's home address (when tied to a profile picture) or the
       | location of research facilities (when images capture experimental
       | equipment)
       | 
       | Oof, that's not too great.
       | 
       | [1] https://github.com/google-research/arxiv-latex-cleaner
        
         | calvinmorrison wrote:
         | They responsibly disclosed it in their research paper. An
         | unethical use would be to use those coordinates to gain state
         | secrets about say, research facilities
        
         | michaelmior wrote:
         | Having arXiv run the cleaner automatically would definitely be
         | cool. Although I've found it non-trivial to get working
         | consistently for my own papers. That said, it would be nice if
         | this was at least an option.
        
         | cycomanic wrote:
         | Leaks of read/write access to documents and GitHub, Dropbox etc
         | credentials is certainly worrying, but location and
         | author/photographer details in photo metadata? That's quite a
         | stretch, and seems like the authors here are just trying to
         | boost the numbers.
         | 
         | The vast majority (I would wager >(100 - 1e-4)) of location of
         | research institutions is public knowledge and can be found out
         | by simply googling the institution address (I am not aware of a
         | single research institution that publishes publically where the
         | location is confidential).
        
       | agarttha wrote:
       | I offer free beer in a comment in my arxiv tex source.
        
       | fcpk wrote:
       | while exif might be bad for private photos, I do think research
       | should not tamper with it unless there is a clear security
       | rationale (ie private photos or things that are meant to o b
       | hidden.. leave the data alone there...
        
       ___________________________________________________________________
       (page generated 2025-10-13 23:01 UTC)