SecurityEngineering/WorkingSessions/09-18-13-NetworkTeam
From MozillaWiki
Details
- Location: MTV-3A All Your Base ( was Fail, but Failed )
- Vidyo: remoties use Sid's vidyo room
- Start time: roughly ~8:45 PDT
- Etherpad: https://etherpad.mozilla.org/network-security-meetup
Agenda and Notes
Introduction
09:00 - Gerv says hi
Kathleen Wilson is the person the person to ask on the CA program in mozilla: How many Certs/orgs, policies, etc.
Doug: Can we disable OCSP now on mobile (android, firefox os).
Doug: we need captive portal
Gerv: things thak make me happy: HSTS, keypinning, cert telemetry, ct, dane, ponies& rainbows. OCSP for EV not for DV? Why not?
- We should have a way to do revo checking before OCSP -> For CA we already do Chemspill.
10:00
DANE?
- I think we all agree it's not the right way forward
- And slow
OCSP
- right now we're saying "OCSP sucks" but not helping improve it
- should we support OCSP GET so CDNs can be used and so we can cache responses?
- Patch in progress: https://bugzilla.mozilla.org/show_bug.cgi?id=436414 -- Gerv
- Must staple: http://tools.ietf.org/html/draft-hallambaker-muststaple-00
- Stamm
- work closer with the Tor project to port in some of their patches so that they can focus on tor and FF can be extended to help them.
- CCH Next Steps...
- No change. No Policy. Sit tight for now.
11:15
HTTP decrypting proxies
- do we want to create a TLS MITM proxy protocol that allows decryption and inspection without requiring addition of a new trust anchor?
- Would be client- (and server-?) authorized
11:30
understand brian's ciphersuite ordering proposal and impacts
- https://briansmith.org/browser-ciphersuites-01.html
- Offer set of ciphersuites that is consistent with other browsers
- _more_ emphasis on ephemeral key exchange
- _more_ emphasis on performance (speed)
- devd is studying cipher suite change impacts (e.g., avoiding RC4, etc)
12:05
Devtools discussion
- FireShark? WireFox? Do we want TLS inspection info in our dev tools?
Brainstorming:
- https://ssllabs.com
- Give advice
- Make it easier to find cipher suite and TLS version used
- Exposing the ephemeral key exchange master secret through devtools for wiresharking. https://developer.mozilla.org/en-US/docs/NSS_Key_Log_Format
- A tool that can generate our recommended OpenSSL cipher suite configuration strings (as used in Apache and other software) (ssllabs)
- Warnings about TLS version intolerance fallback (ssllabs)
- Warnings about lack of TLS 1.2/3 support (ssllabs)
- warning about giant cert sizes (ssllabs can do this)
- Highlight unncessary / wasteful bits of certificates in cert UI and show amount of wasted space. (e.g. Verisign/Symmantec puts their LOGO as a PNG in every cert.) (ssllabs)
- Warn about too-long certificate chains (ssllabs)
- warning about giant cert sizes (ssllabs can do this)
- Warnings about server choice of deprecated ciphersuites (ssllabs)
- Warnings about lack of HSTS / key pinning / must staple (ssllabs)
- Warnings about lack of CT proof in certificate / OCSP response (ssllabs)
- Warnings about missing revocation information (ssllabs)
- TLS record size information (ssllabs)
- Detect vulnerable OpenSSL (etc.) versions through fingerprinting (ssllabs)
- Make it easier to reset cached SSL state / connections
- link to https://ssllabs.com report and be done with things?
- Warnings about badly/wrongly/incompletely constructed certificate chain with a "give me the fixed certificate chain" button that allows them to save the corrected chain. (ssllabs)
- Scan server for cipher suite support (ssllabs)
12:30
LUNCH BREAK
14:00
Goals discussion
- compared/discussed Q3 goals
- SecEng will connect their goals to the platform goals page in Q4 2013
14:40
WiFi Captive Portal detection (562917)
- Firefox team is pretty motivated to make this happen
- Detecting redirects is the first step
15:00
NSPR/NSS branch model stuff
15:40
HTTP/2 IETF stuff
- Should TLS be mandatory for HTTP/2? <- sure.
- Should it mandate ciphersuites?
- Brian: perhaps requiring a minimum bar like TLS 1.1, but maybe this spec isn't the right place to document best ciphersuites.
- Should it mandate ciphersuites?
- We should make a point to develop a plan in Q4 on how to move this forward (together as SecEng and Network)
16:20 Pinning
- Main concern: it's fragile; sites can break.
- Chrome has had Pinning since 2011. Where are we with our pinning?
- Requires insanity
Statistics
- can we do anything around key pinning other than opt-in? Even just for reporting suspected mitm? -- dougt/mcmanus has one dumb idea.
- What if we just report violations or even cert change info and mine that data? (Obvious privacy issues)
- https://bugzilla.mozilla.org/show_bug.cgi?id=846506
Action items:
- 2-second Timeout [camillo - bug 918120]
- disable insecure fallback to SSL 3 [briansmith]
- Captive portal stuff [dougt will find owner] 562917 (lco has done some UI designs, MattN + filipe have thought about this)
- Circulate proposal for cert change system [dougt]
- talk to dcamp about TLS-related dev tools [sid]
- find owner for meta/a referrer stuff [sid - bug 704320]
---
FUTURE Agenda:
- Areas of company focus
- Mozilla-defined NSS future work
- NSPR/NSS Branch Model at Mozilla (false start 713933?)
- Proper Security Reviews
- Cha-Cha evaluation
- Various TLS specifics
- client hello issues - npn/alpn. alpn profiles
- TLS 1.[12'3']
- handshake/IW fit
- Channel ID
CA Policy:
- Why are there so many CA in Firefox?
Future protocol stuff (?):
- Minimum security profile for http/2
- TLS for http:// scheme
- CarpeNet
- state of windows integrated auth
How we can work better together (everyone)