pfg

98 ms·

901.
by pfg 11y ago
It's covered in this blog post[1]. Some mitigations are in place now. [1]: https://blog.torproject.org/blog/tor-security-advisory-relay...
902.
by pfg 11y ago
Why do they specifically mention the fact that the Tor Project warns about possible vulnerabilities? There's a reasonable expectation of privacy in a hotel room, for example[1]. To me this argument reads like: "The hotel owner can
903.
by pfg 11y ago
With a sufficiently complex passphrase, it's not a problem. With a typical 6-digit PIN, disabling the artificial delay and auto-wipe (which is possible via firmware updates) is all you need for a successful brute-force attack.
904.
by pfg 11y ago
To me the most bizarre part of this story is the argument that this didn't constitute an unlawful search because there is no reasonable expectation of privacy for Tor users: > [...] it is the Court's understanding that in order
905.
by pfg 11y ago
IIRC BitLocker with pre-boot authentication mitigates DMA attacks. Most Windows hardware doesn't come with FireWire or Thunderbolt ports nowadays. Microsoft recommends pre-boot auth for devices with DMA ports.
906.
by pfg 11y ago
Their hash algorithm takes ~80ms per attempt. All crypto happens on a dedicated AES engine. Files are encrypted with a key derived from UID (device key which cannot be extracted via firmware) and passphrase.
907.
by pfg 11y ago
Replacing the firmware (Secure Enclave or not) with a version that doesn't include rate-limits and auto-wipe isn't the same as having a crypto backdoor. Right now, your data is still protected by the complexity of your passphrase.
908.
by pfg 11y ago
Yes, via DFU mode.
909.
by pfg 11y ago
The FBI is saying Apple should create a version of their firmware that only works on this specific device. I'm not sure if that's possible, though - if whatever ID they're checking can be spoofed on another device, they'
910.
by pfg 11y ago
I would imagine the firmware signing key is stored on a Hardware Security Module (HSM). Devices like that are tamper resistant (for example, they might be configured to delete the key if someone is messing with the device). HSMs generally s
911.
by pfg 11y ago
This. Telegram got a lot of things right. They have clients that work well on all platforms. Features like the Bot API and Channels have a lot of potential. However, their home-grown crypto and dismissal of feedback from the crypto commun
912.
by pfg 11y ago
I think there are a lot of conflating issues in this discussion. Firmware signing and how updates are delivered are one thing. I would argue that having only one possible adversary is preferable to everyone being able to create firmware tha
913.
by pfg 11y ago
I don't think this is a fair assessment. Apple is being asked to create a firmware version without brute force rate-limiting. With a sufficiently complex passphrase, the FBI is still SOL. They're not being asked to create a firmwa
914.
by pfg 11y ago
First of all, it sets a precedent that companies can be forced to create software that willfully compromises their security mechanisms. This is very much different from your typical data dump ordered by a judge. It's quite likely that
915.
by pfg 11y ago
They're not unusable, it just becomes a bit more inconvenient to use them. In my experience, you need to solve 2-3 captchas per site per new circuit (i.e. every 10 minutes, IIRC, although TFA mentions this might have changed recently).
916.
by pfg 11y ago
Their root certificate is currently not trusted by any browser vendor. They have a cross-sign from IdenTrust, which is trusted even by XP. However, their intermediate certificate uses a name constraint in a way that causes schannel (Microso
917.
by pfg 11y ago
The raw data is publicly available on Certificate Transparency log servers[1]. There are various clients for CT log servers out there, but they all require you to essentially download the entire log to query it (there's no query API).
918.
by pfg 11y ago
FWIW, keeping internal hostnames private might be part of a security-by-obscurity strategy.
919.
by pfg 11y ago
Slightly OT: It would be great if CT logs were available as part of Amazon's or Google's public data sets. Being able to access that data via BigQuery (or similar) would make generating something like this way easier. It would als
920.
by pfg 11y ago
Encrypting sensitive user data is only one advantage of TLS. Without TLS, every single network hop between you and your client could modify your site. This could mean altering the content, adding ads or even malware. Some large ISPs have do
921.
by pfg 11y ago
Your CA should not be in a position to send you your private key. They don't need it in order to sign your certificate. Pretty much every CA I'm aware of allows you to provide your own CSR (which only includes your public key).
922.
by pfg 11y ago
StartSSL and WoSign have been offering free, publicly trusted certificates with one year lifetimes (and the ability to renew for free) for quite some time. The former doesn't allow commercial usage, while the latter operates in China.
923.
by pfg 11y ago
4th largest counting certificates that are either on Certificate Transparency Log servers or in censys' data set. CT logs are populated by CAs sending their certificates to log servers. Only a few CAs (including Let's Encrypt) do
924.
by pfg 11y ago
"any of the public keys in your certificate chain" includes your own certificate, so yes, that's possible. However, if your main and backup key is lost or compromised, you have essentially bricked your domain. That's why
925.
by pfg 11y ago
You can pin to any of the public keys in your certificate chain, so that's the root cert, intermediate certs, or your own cert. Most deployments currently pin to a main and backup CA. This is because HPKP makes it fairly easy to essent
926.
by pfg 11y ago
Every CA in your trust store can issue certificates for every domain. "Attack", in this context, means issuing a trusted certificate that can be used to MitM users of your site. HPKP helps mitigate this risk on a Trust on First Us
927.
by pfg 11y ago
Two CAs already do this (StartSSL, WoSign). FWIW, there's a blog post by CertSimple[1] kind of confirming this is coming (no sources or specifics, though). [1]: https://certsimple.com/blog/domain-validated-ssl-wi
928.
by pfg 11y ago
You don't have to run anything as root if you don't want to. There are tons of clients out there without that requirement. Your argument is basically that Let's Encrypt should have put more focus on working like other CAs do,
929.
by pfg 11y ago
It's my understanding that you can pin to any certificate in your chain, but if a server presents a certificate that leads to a user-defined trust anchor (i.e. your typical corporate MitM proxy cert), no HPKP check is performed.
930.
by pfg 11y ago
Firefox defaults to the same implementation[1]. I'm personally fine with this. MITM proxies are a valid use case. I do think that all parties should be made aware of it, but that's probably something that has to be fixed via laws

More ›