Date: Thu, 23 Jul 2020 09:58:04 -0400
From: Stephen Frost <sfrost@snowman.net>
To: Tal Glo <glozmantal@gmail.com>
Cc: Paul =?iso-8859-1?Q?F=F6rster?= <paul.foerster@gmail.com>,
	Ron <ronljohnsonjr@gmail.com>, pgsql-admin@lists.postgresql.org
Subject: Re: Encryption in pg_dump
Message-ID: <20200723135803.GR12375@tamriel.snowman.net>
References: 
 <CAD7GeHRZ3tu8y9VG_0juVK3RQq6r+t6HciguEz_C8Eh85Hmq3A@mail.gmail.com>
 <2BCA8D4B-E94D-4F2F-99A5-D783912A1CDF@gmail.com>
 <20200722185541.GE3899@momjian.us>
 <6D40812C-3F42-41A2-A548-4EC7754DF473@gmail.com>
 <87zh7qbxqk.fsf@gmail.com>
 <CAJ7S9TW2j7Dk3k62_b_oZg9X=7jKMyLuebcBgtj0aWjSjzGJTw@mail.gmail.com>
 <a91e4430-6e74-2980-b914-7a045d3636ad@gmail.com>
 <4DCA85A3-B307-48E8-9B30-0D8E2E4911D2@gmail.com>
 <CAMreHJLzv9xM2z6vcxq6KxV37C0RGuDSatsa0Fho1u==9i2+Zg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="XPX6T3UXHpsvYx46"
Content-Disposition: inline
In-Reply-To: 
 <CAMreHJLzv9xM2z6vcxq6KxV37C0RGuDSatsa0Fho1u==9i2+Zg@mail.gmail.com>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk


--XPX6T3UXHpsvYx46
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

Greetings,

* Tal Glo (glozmantal@gmail.com) wrote:
> There is a way to implement full homomorphic encryption (FHE) with
> Postgres. The decision which attributes to encrypt must be done at database
> planning step. It worth mentioning that FHE is good to implement, if one
> wants to be able to do server side supported mathematical operations on the
> encrypted data. If the data just needs to be stored in an encrypted form
> then using FHE might be a storage overkill. A dump of the database with FHE
> encrypted attributes preserves the encryption of the encrypted attributes.

Wasn't it relatively recently pointed out that having the ability to do
mathematical operations on the encrypted data ends up leading someone to
be able to deduce a fair bit of information much more easily than
actually breaking the encryption would involve..?

https://eprint.iacr.org/2016/775.pdf

Not finding it right now but I recall there being another paper which
basically pointed out that if you have context information about the
encrypted data (if it's credit card numbers, or SSNs, or such), along
with the ability to perform some operations (addition, ordering, etc)
that you could derive the real values without that much difficulty.

> 1. The encrypted attributes were big in size and Postgres cannot store the
> FHE cypher texts data type. 20,858 encrypted values used 3,6GB. ~33% of
> that size is because of base64 and the rest of the size ist because of my
> chosen encryption parameters.

A bit confused as to why you'd use base64 to store the data- would make
a lot more sense to use bytea.

> 2. Handling queries related to FHE encrypted attributes on the server side
> requires an implementation of own C language functions.

This could certainly be done in an extension, of course.

Thanks,

Stephen

--XPX6T3UXHpsvYx46
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQIcBAEBCgAGBQJfGZdrAAoJEO1sijiDR2RVoisQAMc6epnYZLnUvg18oQLFI9yZ
zG5L4SwfygJCW9QqGK86piy9QMS/s9ylKoXCvPFO8GUBlIKatGiD9LZm22Z8QTD6
snvYzqe5x8rc24qGHzuikyy2lDJllZp5epzNRmh5wXZ4Zjt9sDogN0RTHCft/0vH
Gt061RV6RIk3r915fryliSrBgOjHZ0QwGeBYcfjDco8zddIWfqTvndxcH8awtfIe
E6ojRfHtFTLkcUw/VEIGfltSjH6UHMkx3VxXSwkMxSNvTiRFVHrk8DMhqoClY2bg
MQbx8mEiaQd3PvDOFcctIi/u0BavbnV+4g5G6c0N3JeeF2ROKck+S6Ntq8te3Toa
3qxxTD9THbLpxgLLDI5JB9Ie4SSFFmhbUidtmNoXIABbyhCo3Id4RwsHlH3r3PC8
WDYf3kjzfruZPCnjkx6BFHxbAm06Pa4EA9YlJqfFoVVaNGJ5tf5AVLFuMzqEAZVl
zlCiaTkxHK8hy6VUgVXBDwOYPPdf+35qrepn9dwB5FYK9scwNKbD1BpN4XaW6TVO
8rhu2TooSpfin6jktKe6W3f7J7vguKrgvm+Ig2uLYUyeWhMj6gV4w0/445SHtdH8
BTSWcsXmq66AaLRrzZA3C31LSbJabTia6o7cISegBXKvEpkWC7u77b+2Lvh44XeG
uAUSICg860S4rQNAtYOY
=l3jE
-----END PGP SIGNATURE-----

--XPX6T3UXHpsvYx46--