Re: [openpgp] Mitigation of Attacks on Email End-to-End Encryption

[Marcus Brinkmann <[e-mail]>]

Hi,

sorry, I did not test the DOI. The URL is:

https://dl.acm.org/doi/10.1145/3372297.3417878

The PDF is at:

https://dl.acm.org/doi/pdf/10.1145/3372297.3417878

Thanks,
Marcus

On 11/3/20 5:35 PM, Stephen Farrell wrote:
> 
> Hiya,
> 
> Interesting work and I look forward to seeing your
> suggestions as we go.
> 
> One thing: the DOI URL below gets me a "not found"
> error but indicates that may change in future. Is
> that correct? If so, be great if you could send a
> ping when the paper is available (or another URL
> to a preprint or whatever).
> 
> Thanks,
> S.
> 
> 
> On 03/11/2020 16:24, Marcus Brinkmann wrote:
>> Hello,
>>
>> we have just published a paper on our research how to mitigate attacks
>> on email end-to-end encryption. The full paper is available here with
>> open access:
>>
>> Jörg Schwenk, Marcus Brinkmann, Damian Poddebniak, Jens Müller, Juraj
>> Somorovsky, and Sebastian Schinzel. 2020. Mitigation of Attacks on Email
>> End-to-End Encryption. In Proceedings of the 2020 ACM SIGSAC Conference
>> on Computer and Communications Security (CCS '20). Association for
>> Computing Machinery, New York, NY, USA, 1647–1664.
>> DOI:https://doi.org/10.1145/3372297.3417878
>>
>> We have analyzed three attack vectors: EFAIL malleability gadgets (MG),
>> EFAIL direct exfiltration (DE), and REPLY attacks.
>>
>> MG attacks exploit unauthenticated block cipher modes such as CBC and
>> CFB, and are mitigated by using an authenticated encryption mode such as
>> AEAD, or by a strict implementation of OpenPGP's modification detection
>> code.  S/MIME 4.0 and OpenPGP RFC4880bis have added AEAD encryption
>> modes, and assuming a strict implementation, can also protect against MG
>> attacks that way. (Currently, RFC4880bis does allow unsafe
>> implementations of AEAD, and even encourages them due to unrestricted
>> chunk sizes. This has been previously discussed, and I will re-raise
>> this issue when the WG has been reinstantiated).
>>
>> EFAIL DE attacks rely on modifications of the MIME structure to embed
>> authentic ciphertexts in a context that allows exfiltration of the
>> plaintext after decryption, for example through image source URLs in a
>> HTML MIME element before the ciphertext. These attacks have also been
>> published in the EFAIL paper, and so far were only mitigated at the
>> recipient side by ad-hoc measures in email clients. Our experience was
>> that developers were struggling to mitigate these attacks. For example,
>> we found several bypasses after attempts at mitigation.
>>
>> REPLY attacks are known for 20+ years: They rely on modifications of the
>> email header (SMTP) context, that allow the attacker to receive replies
>> to authentic ciphertexts, were the victim quotes the plaintext back to
>> the attacker. To our knowledge, these attacks have not been mitigated so
>> far. We have looked at reply attacks in our paper on covert content
>> attacks [COVERT].
>>
>> We have looked systematically at these issues, and propose to protect
>> the MIME and SMTP context of an email by adding a summary of this
>> decryption contexts (DC) as associated data (AD) in the AEAD encryption.
>> This way, any significant modification to these contexts that indicate
>> an attack would lead to a decryption error, rather than emitting the
>> plaintext to the application, where it would be subject to a large
>> attack surface to launch DE or REPLY attacks.
>>
>> To support this mechanism, OpenPGP RFC4880bis would need to be amended
>> to allow applications to add arbitrary data to the AD, either directly
>> (length+value) or by adding a hash representation (constant length). I
>> plan to introduce a proposal for these changes when the WG is
>> reinstantiated.
>>
>> We have evaluated which SMTP headers are relevant for REPLY actions in
>> email clients, and which MIME contexts can be considered safe. Based on
>> this (and inspired by DKIM), we make a specific proposal for calculating
>> the decryption context which is sender-enforced and extensible.
>>
>> We have implemented this solution with GnuPG and Thunderbird/Enigmail.
>> It was easy to implement, offered excellent compatibility (low false
>> positive rate when detecting attacks) and mitigated all REPLY and DE
>> attacks conclusively without introducing new cryptographic primitives.
>>
>> I hope these findings are interesting to the OpenPGP community. For
>> example, some of the problems described in
>>
>> https://www.ietf.org/archive/id/draft-dkg-lamps-e2e-mail-guidance-00.txt
>>
>> can be mitigated using our techniques. Also, the REPLY action behavior
>> of email clients in our evaluation should be useful for the memory hole
>> project, for example.
>>
>> Thanks,
>> Marcus
>>
>>
>> Bibliography:
>>
>> [COVERT] Jens Müller, Marcus Brinkmann, Damian Poddebniak, Sebastian
>> Schinzel, and Jörg Schwenk. 2019. Re: What’s Up Johnny? – Covert Content
>> Attacks on Email End-to-End Encryption.
>> https://arxiv.org/ftp/arxiv/papers/1904/1904.07550.pdf.
>>
>> [EFAIL] Damian Poddebniak, Christian Dresen, Jens Müller, Fabian Ising,
>> Sebastian Schinzel, Simon Friedberger, Juraj Somorovsky, and Jörg
>> Schwenk. 2018. Efail: Breaking S/MIME and OpenPGP Email Encryption using
>> Exfiltration Channels. In 27th USENIX Security Symposium, USENIX
>> Security 2018, Baltimore, MD, USA, August 15-17, 2018., William Enck and
>> Adrienne Porter Felt (Eds.). USENIX Association, 549–566.
>> https://www.usenix.org/conference/usenixsecurity18/presentation/poddebniak
>>
>>

-- 
Dipl.-Math. Marcus Brinkmann

Lehrstuhl für Netz- und Datensicherheit
Ruhr Universität Bochum
Universitätsstr. 150, Geb. ID 2/461
D-44780 Bochum

Telefon: +49 (0) 234 / 32-25030
http://www.nds.rub.de/chair/people/mbrinkmann