[Haskell-cafe] How to use an crypto with hackage cereal?

Wed Apr 22 03:41:54 UTC 2015

That is the ugliness of the original binary data. The encryption is not by
fixed block size. So decrypt cannot be run before the get* helpers.
So decrypt-runGetPartial-decrypt-runGetPartial loop would not work.

I need a "post process" in Get. For example, "portNumber <- liftM decrypt
getWord16be; return $ MyDataType portNumber". But currently I could not
pass decrypt into get function.

On Wed, Apr 22, 2015 at 11:26 AM, Andrey Sverdlichenko <blaze at ruddy.ru>
wrote:

> You can't really modify source bytestring inside Get monad, and this is
> what decryption effectively do. The only option I know about is to run
> another parser inside Get monad. I'd rather write
> decrypt-runGetPartial-decrypt-runGetPartial loop and return Fail from it on
> decryption error.
>
>
>
> On Tue, Apr 21, 2015 at 8:12 PM, Magicloud Magiclouds <
> magicloud.magiclouds at gmail.com> wrote:
>
>> How about fail in Get monad if decrypt failed? So decrypt failure would
>> lead to a result of "Left String" on decode.
>>
>> On Wed, Apr 22, 2015 at 11:05 AM, Andrey Sverdlichenko <blaze at ruddy.ru>
>> wrote:
>>
>>> You probably should not merge decrypt and decode operations, it is bad
>>> crypto habit. Until you decrypted and verified integrity of data, parsing
>>> is dangerous and opening your service to attacks. Correct way of
>>> implementing this would be to pass ciphertext to decryption function and
>>> run parser only if decryption is successful. If bytestring is too big to be
>>> decrypted in one piece, consider encrypting it in blocks and feeding
>>> decrypted parts to parser.
>>>
>>>
>>>
>>> On Tue, Apr 21, 2015 at 7:49 PM, Magicloud Magiclouds <
>>> magicloud.magiclouds at gmail.com> wrote:
>>>
>>>> Similar as you envisaged. I would receive a bytestring data and a
>>>> config point out what cipher to use. Then I deserialize the data to a data
>>>> type with some fields. The serialize process is something like:
>>>>
>>>> msum $ map (encrypt . encode) [field1, field2, field3]
>>>>
>>>> I could parse the bytestring outside Get/Put monads. But I think that
>>>> looks ugly. I really want to embed the decrypt process into Get/Put monads.
>>>>
>>>> On Tue, Apr 21, 2015 at 10:08 PM, Ivan Lazar Miljenovic <
>>>> ivan.miljenovic at gmail.com> wrote:
>>>>
>>>>> On 21 April 2015 at 23:58, Magicloud Magiclouds
>>>>> <magicloud.magiclouds at gmail.com> wrote:
>>>>> > Thank you. But how if the cipher was specified outside the binary
>>>>> data? I
>>>>> > mean I need to pass the decrypt/encrypt function to get/put while
>>>>> they do
>>>>> > not accept parameters. Should I use Reader here?
>>>>>
>>>>> Maybe you could explain what you're doing better.
>>>>>
>>>>> I would envisage that you would get a Bytestring/Text value, then
>>>>> encrypt/decrypt and then put it back (though if you're dealing with
>>>>> Bytestrings, unless you're wanting to compose them with others there's
>>>>> no real need to use Get and Put as you'll have the resulting
>>>>> Bytestring already...).
>>>>>
>>>>> Or are you wanting to implement your own encryption/decryption scheme?
>>>>>  In which case, you might want to either:
>>>>>
>>>>> a) write custom functions in the Get and Put monads OR
>>>>>
>>>>> b) write custom parsers (e.g. attoparsec) and builders (using the
>>>>> Builder module in bytestring); this is probably going to suit you
>>>>> better.
>>>>>
>>>>> >
>>>>> > On Tue, Apr 21, 2015 at 6:43 PM, Yitzchak Gale <gale at sefer.org>
>>>>> wrote:
>>>>> >>
>>>>> >> Magicloud Magiclouds wrote:
>>>>> >> > I am trying to work with some binary data that encrypted by field
>>>>> >> > instead of
>>>>> >> > the result of serialization. I'd like to use Data.Serialize to
>>>>> wrap the
>>>>> >> > data
>>>>> >> > structure. But I could not figure out how to apply an runtime
>>>>> specified
>>>>> >> > cipher method to the bytestring.
>>>>> >>
>>>>> >> Are you using the set of crypto libraries written by
>>>>> >> Victor Hanquez, such as cryptocipher-types,
>>>>> >> crypto-pubkey-types, and cryptohash?
>>>>> >>
>>>>> >> Or the set of libraries written by Thomas DuBuisson,
>>>>> >> such as crypto-api, cipher-aes128, etc.?
>>>>> >>
>>>>> >> Here is an example of decoding for Victor's libraries.
>>>>> >> Encoding would be similar using Put instead of Get.
>>>>> >> Thomas' libraries would be similar using the other
>>>>> >> API.
>>>>> >>
>>>>> >> Let's say you have a type like this:
>>>>> >>
>>>>> >> data MyCipher = MyAES | MyBlowfish | ...
>>>>> >>
>>>>> >> Then in your cereal code you would have a Get monad
>>>>> >> expression something like this (assuming you have
>>>>> >> written all of the functions called parseSomething):
>>>>> >>
>>>>> >> getStuff = do
>>>>> >>   cipher <- parseCipher :: Get MyCipher
>>>>> >>   clearText <- case cipher of
>>>>> >>     MyAES -> do
>>>>> >>       keyBS <- parseAESKey :: Get ByteString
>>>>> >>       let key = either (error "bad AES key") id $ makeKey keyBS
>>>>> >>           cipher = cipherInit key
>>>>> >>       cipherText <- parseAESCipherText :: Get ByteString
>>>>> >>       return $ ecbDecrypt cipher cipherText
>>>>> >>     MyBlowfish -> do ...
>>>>> >>
>>>>> >> etc.
>>>>> >>
>>>>> >> Hope this helps,
>>>>> >> Yitz
>>>>> >
>>>>> >
>>>>> >
>>>>> >
>>>>> > --
>>>>> > 竹密岂妨流水过
>>>>> > 山高哪阻野云飞
>>>>> >
>>>>> > And for G+, please use magiclouds#gmail.com.
>>>>> >
>>>>> > _______________________________________________
>>>>> > Haskell-Cafe mailing list
>>>>> > Haskell-Cafe at haskell.org
>>>>> > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
>>>>> >
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Ivan Lazar Miljenovic
>>>>> Ivan.Miljenovic at gmail.com
>>>>> http://IvanMiljenovic.wordpress.com
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> 竹密岂妨流水过
>>>> 山高哪阻野云飞
>>>>
>>>> And for G+, please use magiclouds#gmail.com.
>>>>
>>>
>>>
>>
>>
>> --
>> 竹密岂妨流水过
>> 山高哪阻野云飞
>>
>> And for G+, please use magiclouds#gmail.com.
>>
>
>

-- 
竹密岂妨流水过
山高哪阻野云飞

And for G+, please use magiclouds#gmail.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20150422/d81c0d06/attachment.html>