Stories
Slash Boxes
Comments

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

Encrypted But Searchable Online Storage?

Posted by timothy on Thu Apr 16, 2009 03:20 PM
from the give-some-to-that-lawyer dept.
An anonymous reader asks "Is there a solution for online storage of encrypted data providing encrypted search and similar functions over the encrypted data? Is there an API/software/solution or even some online storage company providing this? I don't like Google understanding all my unencrypted data, but I like that Google can search them when they are unencrypted. So I would like to have both: the online storage provider does not understand my data, but he can still help me with searching in them, and doing other useful stuff. I mean: I send to the remote server encrypted data and later an encrypted query (the server cannot decipher them), and the server sends me back a chunk of my encrypted data stored there — the result of my encrypted query. Or I ask for the directory structure of my encrypted data (somehow stored in my data too — like in a tar archive), and the server sends it back, without knowing that this encrypted chunk is the directory structure. I googled for this and found some papers, however no software and no online service providing this yet." Can anyone point to an available implementation?
+ -
story

Related Stories

This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • by nahdude812 (88157) * on Thursday April 16 2009, @03:25PM (#27602719) Homepage

    It's not possible to do this even in theory, unless you're relying on very weak encryption. The point of encryption is that you can't infer anything about the contents. If Google was able to infer enough to give you meaningful search results (if for example each word was encrypted by itself, and you searched for the encrypted version of the word), they would therefore necessarily be able to know enough to perform a frequency analysis attack on your data and compromise it in no time flat unless it was a very small amount of data (thus meaning search isn't really of value anyway).

    You'll find a similar problem plagues any attempt at searching. Searching requires a certain knowledge or meta knowledge of the material being searched; and that knowledge necessarily dramatically weakens your encryption.

    • by TheRaven64 (641858) on Thursday April 16 2009, @03:28PM (#27602773) Homepage Journal

      It is possible. When you upload the data, you also upload an index. When you connect again, you download the index (which is much smaller than the data) and search that on the local machine. Neither the index, nor the data, is ever unencrypted on the server.

      As for frequency analysis, I don't think any encryption algorithms published in the last 40 years have been vulnerable to this sort of attack...

      • That's because all encryption produced in the last 40 years has been based off of Division not Addition
      • by TheRaven64 (641858) on Thursday April 16 2009, @03:34PM (#27602861) Homepage Journal
        Replying to myself: the scheme in the linked paper is not feasible. It performs O(n) searches, but this means that the amount of data you need to upload for the query is equal to the total amount stored. Since most consumer Internet links are asymmetric, it would be cheaper and easier to simply download the entire data search locally. The paper proposes having a server-side cache. This means that, for a typical block cypher, you would have a cache of every search term encrypted for each block. The server could then compare this to each block, but would not know what the plaintext is. This is not useful in any real-world scenario. The cache would be orders of magnitude bigger than the stored data and the search would sill be O(n), which is painfully slow. As I suggested above, uploading an encrypted index with the data makes more sense. Look at Apache Lucene or Apple's SearchKit for how to do this.
        • Re: (Score:3, Insightful)

          > the amount of data you need to upload for the query is equal to the total amount stored
          That's not how I read it. But the approach still sounds useless:

          If Alice wants to search for the word W, she can tell Bob (the server) the word W and the ki corresponding to each location I in which W may occur

          What's the use of encrypting the data if you're going to send keywords in cleartext to a party you're trying to hide the data from?

      • by FredFredrickson (1177871) * on Thursday April 16 2009, @03:34PM (#27602865) Homepage Journal
        Mozy does this for personal/business backups. You can use a completely private key, but search your own data.
      • Re: (Score:3, Informative)

        by BitZtream (692029)

        And that would practically defeat the purpose of the encryption.

        For the index to be useful it has to provide too much information about the encrypted data. The point of encryption is to ensure that nothing can be inferred about the contents of the encrypted data. If you give them a nice big bunch of information about whats encrypted, why bother encrypting it in the first place?

        Given enough information in the index they could actually derive your encryption key as well with some simply brute forcing.

    • Unless you do the indexing client-side, and upload an index that's somehow encrypted...

      I'm not saying I know how to do this, but it seems possible.

      • Re:Maybe, maybe not (Score:4, Interesting)

        by The Moof (859402) on Thursday April 16 2009, @03:42PM (#27603035)

        Maybe something like this -

        Create an index of hashes using the unencrypted data on the client.
        Encrypt the data on the client so we now have an index of hashes that apply to an encrypted file.
        Upload the hash index and the encrypted data file to the server.
        To search, hash the search criteria on the client.
        Server search the indexes for the hash value, returning a list of encrypted files with an index matching the criteria hash.

    • by blueg3 (192743)

      Not possible in theory? You should tell the authors of the linked paper that describe how to do it in theory.

      • The algorithm in the linked paper requires you to upload at least as much data as is stored remotely for every search query. This is technically possible, but it would be cheaper and easier to download and decrypt all of the data locally then run all of your searches, which seems to defeat the point. The only occasion when their algorithm makes sense is when you are repeatedly searching for the same terms, but if you're doing that then you should just save your search results.
    • Yeah, I'm not sure I understand how meaningful searches can be done without decryption-- but then I don't pretend to be any kind of a genius about these things. It seems much more likely to me that there could be some kind of a system where unencrypted search indexes are kept locally while the files are encrypted and sent to an online storage service. Then you could search locally for the file you're looking for, fetch the encrypted information from the online storage, and then decrypt it locally.

      That so

    • by smallfries (601545) on Thursday April 16 2009, @03:40PM (#27602973) Homepage

      I'm curious - why would you post a comment claiming that this can't even be done in theory, when the submitter included links in the summary to a paper that shows that it can?

      • Re: (Score:3, Interesting)

        by raju1kabir (251972)

        I'm curious - why would you post a comment claiming that this can't even be done in theory, when the submitter included links in the summary to a paper that shows that it can?

        Because the paper doesn't propose any solution that is practical, or which even leads to a practical solution.

        In theory I can cure all forms of cancer - all I have to do is go through each cell in the victim's body and pluck out the cancerous ones.

    • There are encryption algorithms that allow addition. That is, the sum of two encrypted messages is an encryption of the sum. I've forgotten how these work exactly, I think they are some many to one mapping, and the addition operation is not simply adding the encrytped numerical representations.

      I came across these when looking at voting systems that allow N distributed people to vote in a way that sums the result before it is decrypted rather than decrypting to do the sum.

      Anyhow what this means is that is

    • by goodmanj (234846) on Thursday April 16 2009, @04:34PM (#27603877)

      Can I have an anti-theft system for my car, so that nobody can steal it but anybody who wants to can take it for an anonymous test-drive?

  • You want to... (Score:4, Insightful)

    by mhkohne (3854) on Thursday April 16 2009, @03:25PM (#27602725) Homepage

    Use an encrypted query to match against the encrypted text. The problem is, if the text is REALLY encrypted, then there shouldn't be enough information to do this - the encrypting of the original text should make it impossible to even match against it.

    If it didn't, then an attacker who got hold of the encrypted text and some of your encrypted queries might well be able to mount an attack based on commonalities between the two.

    • by noidentity (188756) on Thursday April 16 2009, @04:27PM (#27603777)

      The problem is, if the text is REALLY encrypted, then there shouldn't be enough information to do this - the encrypting of the original text should make it impossible to even match against it.

      NOT TRUE! I use a combination of XOR and rot-13 encryption and I'm able to do text searches just fine. The trick is to encrypt the search string, then it'll work perfectly. This is because the encryption doesn't depend on the position within the text, but that shouldn't hurt security too much.

  • This sounds pretty easy,
    a) obtain database, indexing tools, search tool
    b) install on the machine and encrypt the entire hard drive with any of the many available whole-disk encryption tools
    c) ssh in and run queries.

  • Just to clarify the OP's idea. They want to store only encrypted data on the server, send only encrypted queries to the server(that the server can't even decrypt), yet they expect that the server will be able to send them back results. I don't think it can happen but surprise me.

    The best I think you can do is store and transfer the data in encrypted form and put the indexes and any search logic on the client. Maybe the index could be stored on the server as well and synced to the client, but creating the
  • by davidwr (791652) on Thursday April 16 2009, @03:32PM (#27602827) Homepage Journal

    If the data is encrypted in independent "chunks" from which search terms can be built then this is trivial: You pre-encrypt your search terms and search for them. Searching a word ROT13 [wikipedia.org]-encoded document works this way, as each character is encrypted individually and an encrypted search term is made up of encrypted characters.

    Once you get past this, it's no longer easy. You basically have to either make the term you are searching for look like all possible values of the encrypted text and return all matches, or decrypt the document somewhere.

    If the encryption is good and any particular chunk, extract, or other slicing-and-dicing of the encrypted data without the key looks random, you are pretty much stuck with decrypting it somewhere.

    The alternative is to store an index, or at least a list of keywords, in clear text. For example, a document describing how to build a nuclear bomb could have a list of 10 or 20 non-classified keywords attached to it to aid searching. But that's not what you are asking for.

  • There are techniques to do this but none have made it out of academia. Most are quite inefficient and support very restricted querying models. Here's one paper that claims their methods are "practical" (but always keep in mind that academic claims of practicality should always be taken with a grain of salt):

    http://www.cs.berkeley.edu/~dawnsong/papers/se.pdf [berkeley.edu]

  • It's been done. GNUnet [gnunet.org].

  • by skathe (1504519) on Thursday April 16 2009, @03:37PM (#27602915)
    ...and when the bartender asks him what he would like to drink, the guy says "I want what I always get, but I don't want you to actually pour the drink, just help me search behind the bar for the liquor I want, and the hand it to me without seeing what it actaully is, and charge me correctly without any knowledge of what it is you just helped me find."
    • Re: (Score:3, Interesting)

      by richie2000 (159732)

      But... That's not a valid car analogy since you're not allowed to drink and drive.

    • by HTH NE1 (675604) on Thursday April 16 2009, @03:59PM (#27603363)

      Not good enough. The bartender could audit his liquor to see how much of each bottle was dispensed.

      This is why when they do this sort of thing, the gentleman just serves the bartender a National Security Letter and takes more than what he wants without paying a dime.

    • Re: (Score:3, Interesting)

      by dimeglio (456244)

      ...if I tell you a story in French and you don't understand it, you will have no idea what I told you and will not be able to answer questions about my story. However, if you are able to memorize all I told you phonetically I can ask if I said a word or not just by the sound. Yet you don't know exactly what I asked for, nor the meaning of the answer but you are able to answer that question since it doesn't imply meaning.

      So a possibility for the OP would be to store the information in a language unknown to a

  • by Lord Ender (156273) on Thursday April 16 2009, @03:37PM (#27602919) Homepage

    Keep the files on the remote server, encrypted. Keep the search index in a database, encrypted in chunks. Rsync your search database between your local machine and the server. Actual searches of the databases would be done locally.

    Result: terrible performance whenever you access your data from a new machine (must sync entire search database). Good performance the rest of the time. Remote server never sees anything but cyphertext.

  • by dschuetz (10924) <slash@david.das[ ].org ['net' in gap]> on Thursday April 16 2009, @03:41PM (#27603011) Homepage

    ...isn't this easy?

    Plaintext: "Attack at dawn"
    Ciphertext: "lkaoiuast98u;aw"
    Search query: "oiua"
    Result: "lkaoiuast98u;aw"

    What could be simpler?

    (no, I'm not an idiot, this is a joke.)

  • by burnin1965 (535071) on Thursday April 16 2009, @03:45PM (#27603085) Homepage

    As long as your query looks something like this...

    SELECT * FROM mydata WHERE stuff LIKE '%YToyOntzOjc6InBhY2thZ2UiO3M6MjM5OiKyKHPh9ZawDX6KyA62cMd6p+mjBybGwJyCaNfFb7S.........

    Seriously though, if I understand your objective I think it would be feasible to develop something like that, but I don't think its something you could integrate into Google's search services unless they added something on their end.

    You could pass a decryption key along with your query and the server would then decrypt records as it performed the search. It would be very resource intensive.

    As an close example, I have a web based password storage application in which I did not want to keep the encryption keys on the same server as the password database. So I generate a key with which to encrypt the records and the user keeps their key and must supply it every time they want to decrypt a record. I don't go so far as to enable searching of the encrypted data, I have a description field specifically for that purpose. The web application is called Passbox [sourceforge.net] and is written in PHP.

  • What an oxymoron! (Score:3, Interesting)

    by hesaigo999ca (786966) on Thursday April 16 2009, @03:47PM (#27603113) Homepage Journal

    Yeah, Id like my cake and eat it too!

    The only way this could work is if you has tags in the meta header of the encrypted file
    telling you that yes I am encrypted, but I have an image in me or my encrypted data is of the type accounting.

    This might work for indexing searches where you want to be able to return all the files on the pc (encrypted or not) that are images or etc...

  • by Naerbnic (123002) on Thursday April 16 2009, @04:06PM (#27603475)

    There is a cryptography technique called Public Information Retrieval which allows you to do just that: Send an encrypted query to a server, let it perform some operations on your behalf, and send you an encrypted query result. The server neither knows the contents of the encrypted data, nor the content of the query, but you have your result nonetheless.

    The intuition is that there exists a sort of "black-box" operation which some cryptographic techniques can use. For example, if I have two encrypted bits a and b (where I can't tell what a and b actually are), I can still perform the operation a xor b. The result is encrypted, and I don't know the actual operands or the result, but I know that what came out is indeed the encryption of the xor of the encrypted bits. Such cryptosystems are forms of "Homomorphic Encryption".

    Using this, we can then give the server a search term thus encrypted and, using the black-box opertaion, have it do some set of operations which will reveal the result. The server will execute the exact same set of operations independent of the search term, so it knows nothing (and needs to know nothing) of the search term contents. Of course, this implies that the server has to operate on every element of the encrypted data to do its job, but that's the fundamental tradeoff. If you're willing to accept that, and the additional computational overhead, you can design such a system.

  • by airjrdn (681898) on Thursday April 16 2009, @04:30PM (#27603821) Homepage
    But it may not be everything you're looking for. My requirements were:
    1 - Mask the filename
    2 - Encrypt the contents
    3 - Add recovery data in case the file got damaged
    4 - Ability to view unmasked filename from web

    I put together a batch file I could drag/drop multiple files onto that used WinRAR to compress the files (individually), with encrypted filenames, a password (of course), and included archive recovery data. It then used ReNamer to encrypt the .rar filenames. After that, I simply FTP'd the files to the server.

    I had a webpage that would accept a password, and unencrypt the filenames so they were viewable in readable form on the page. Each one was a hyperlink. There was an extra step required if you wanted the downloaded filename to be unencrypted as well.

    After uploading 115G or so, my host alerted me to the fact that they didn't allow me to keep offsite backups there. :) So in the end, I'm not even using it at the moment.

    My solution didn't allow me to search within the files, but it did allow me to store files on the server that they had no way of viewing the contents of, or guessing the contents of based on filename.
    • by qbzzt (11136) on Thursday April 16 2009, @03:26PM (#27602729)

      You're missing something. SSL is for data that is in transit. The poster wants the data to be encrypted on the server. That's easy - any encryption program can do it. But then s/he also wants to search it. That is harder.

        • by vidarh (309115) <vidar@hokstad.com> on Friday April 17 2009, @12:57AM (#27608067) Homepage Journal
          No, it's not impossible. It's not even particularly hard. You do have some limitations though:

          Search works by tokenizing a document, and creating an inverse index from tokens to documents. The tokens does not need to mean anything to the search engine. If you generate the tokens on the client, and don't transmit the dictionary that maps from word to token id, you can have "encrypted search".

          The problem with doing that directly is that if you want to do proximity based search you need information on the token order, and they could do frequency analysis to come up with plain text guesses if they guess the language right. You can counteract that by mapping the same word to multiple tokens to even out the frequency of each token id, but it means you would need to search for multiple tokens to find all occurrences of a word.

          If you don't are about word proximity it's much safer, as the index would only contain each token once per document at most.

          • Re: (Score:3, Insightful)

            by seifried (12921)
            And these tokens are generated how? Oh yeah. by Google's search engine. Whoops. If you want someone to extract information from data they will be definition be able to extract some amount of information from the data, even you have everything encrypted/etc. they could do frequency counts of the tokens and convert them to words, traffic analysis (can't encrypt the from/to). etc.
    • by 3p1ph4ny (835701) on Thursday April 16 2009, @03:30PM (#27602793) Homepage

      No, this is not what SSL is for at all. SSL you have a party you wish to communicate with, but an insecure channel.

      Here, you don't want to communicate anything useful to anyone. This is more a privacy preserving data mining problem. It goes something like this:

      I have a long list of secret numbers 1...n. I do something to these numbers, so that Google doesn't know what they are, and then I send them to Google. Next, I want to know how many numbers are larger than, say k. So, I ask Google, but in a clever way, so that Google doesn't know what I'm asking.

      Google then tells me how many of my original numbers were larger than k. However, Google doesn't know my original numbers, and they don't know what question I asked. There needs to be some theoretical mapping that preserves this privacy, but still allows the data mining to occur.

      • Re: (Score:3, Interesting)

        What it sounded like was that he wanted to keep a database with Google that was encrypted and wanted to search it remotely and securely, but without Google being able to look at the data. Even if that were possible, why are you trusting Google with that in the first place? Why not store it somewhere else? I would think keep encrypted data on a server and make a secure connection to it. You send your normal query across the encrypted channel to the secure server, it does its regular search and sends the resu
        • by jcwayne (995747) on Thursday April 16 2009, @04:59PM (#27604225)

          It can't, that's why I use Live Search. It doesn't understand the query, the data, or the result. Unfortunately, for the OP, it doesn't support encryption.

        • Re: (Score:3, Insightful)

          by vidarh (309115)
          Here's a search index: [1,55] [2,103] [2,178] [3,1] [3,2]. Give me all documents with a document id matching the second entry in each pair where the first entry is 2.

          Do you know which word "2" represents, or what is in documents 103 and 178?

          That's how you do it. You need to ensure there's no way of doing statistical analysis on the token list to recover plaintext info, and you need to not give them the dictionary mapping from plaintext to tokens.

    • Re: (Score:3, Insightful)

      by deroby (568773)

      Yes you are =)

      SSL only encrypts the transport.

      It seems that the poster wants to have his data _stored_ in an encrypted way that is only decipherable by him, not by any of the machines/users at the storage facility. Yet, when he wants to do some search, he somehow expects the server to be able to do so... AFAIK that's not feasible.

      (you could store whatever encrypted stuff remotely, but querying will require fetching, reading and decrypting the (relevant portions of) data locally...)

          • Re: (Score:3, Interesting)

            by mcrbids (148650)

            Sure the *NAME* is "Secure Sockets Layer", and perhaps that was what it was originally developed for, but it's just wrong to say that it can't be used otherwise, and/or that it only encrypts data "in transit", not on a server. Take a look at this:

            http://us2.php.net/manual/en/function.openssl-public-encrypt.php [php.net]

            Here's the use of SSL functionality without (ahem) a socket. Right from the docs:

            This function can be used e.g. to encrypt message which can be then read only by owner of the private key. It can be als

    • Re:huh? (Score:5, Funny)

      by HTH NE1 (675604) on Thursday April 16 2009, @03:37PM (#27602921)

      if the server cannot decipher the query it cannot execute it on a binary blob of encrypted data. FAIL.

      Gung jbhyq qrcraq ba ubj gevivny lbhe rapelcgvba zrgubq vf.

    • Re:huh? (Score:5, Insightful)

      by oldspewey (1303305) on Thursday April 16 2009, @03:38PM (#27602955)

      Well that depends whether the OP wants to perform something like a fulltext search (i.e. the ability to look for keywords within the content of each document) or a metadata search.

      There's nothing to prevent you setting up a CMS where each piece of content is encrypted, but the metadata describing that content is out in the clear and searchable. Security in such a scenario would be less than optimal (e.g. people could guess certain things about your content based on the statistical pattern of length for each of the millions of encrypted content items), and of course you'd have to be very careful about the metadata fields and how you are populating them.