Why TLS is better without STARTTLS ...
summary
Structured testing with the semi-automatic EAST toolkit finds over 40 STARTTLS-specific issues by analyzing 28 clients and 23 servers, with only 3 clients and 7 servers showing no STARTTLS flaws. Attack classes include negotiation stripping, command and response injection via buffering, session fixation, UI spoofing and tampering; credential theft and cross‑protocol HTTPS attacks are demonstrated. Internet scans show about 2% of 60 million servers still vulnerable to the decade‑old command‑injection bug. Recommendation: move to implicit TLS or fully isolate and minimize the plaintext negotiation phase.
AI-written summary · about
chapters (10)
- 0:00Email security and TLS
- 2:29STARTTLS problems
- 9:26Research methodology
- 12:14STARTTLS negotiation attacks
- 16:24Session fixation flaws
- 19:57Command injection flaws
- 26:17Scanning results
- 29:51STARTTLS fixes
- 33:03Vulnerability disclosure challenges
- 35:32Conclusion
related talks
19mWhat's happening in TLS (transport layer security)?
53mHansel and Gretel do TLS
40mWebsite security 101
53mThe BEAST Wins Again: Why TLS Keeps Failing to Protect HTTP
49mAnalysis of DTLS Implementations ... State Fuzzing
01h 01mHTTPS & TLS in 2016: Security practices from the front lines
37mHTTPS Is Better than Ever Before. Now Its Your Turn.
56mSSL/TLS for Mortals