|
LeftistMuslimObama posted:seems more like code injection op
|
# ? Feb 21, 2017 01:11 |
|
|
# ? May 18, 2024 18:54 |
|
pr0zac posted:i'm kinda confused what exactly the point you're trying to make is? the point was originally that sata passwords are probably good enough for people running windows home user editions, which do not come with bitlocker FDE, and who do not want to have to rely on third party FDE. any important machine would of course at least have a pro edition, and thus have bitlocker built in, negating a need to bother with sata passwords.
|
# ? Feb 21, 2017 01:14 |
|
Jewel posted:continuing on from the bitcoin CI, a great one i just saw on twitter lol conceptually similar to the eyepyramid vm detection; should we just call environmental detection and subsequent behavior modification "vdubbing" from now on?
|
# ? Feb 21, 2017 01:29 |
|
Midjack posted:should we just call environmental detection and subsequent behavior modification "vdubbing" from now on? I like this idea
|
# ? Feb 21, 2017 02:19 |
|
Do physical security fuckups qualify for this thread?quote:A group of 11 people went through an unmanned TSA security gate at John F. Kennedy International Airport and are now believed to have boarded flights, according to NBC News investigative reporter Tom Winter, citing officials. quote:Winter said on Twitter at least three of the 11 people set off the metal detector when going through the gate, as seen on a review of police surveillance footage. quote:Sources tell NBC News, TSA officials, "did not notify the Port Authority Police until two hours after breach occurred." ... When they were finally notified, Port Authority cops flooded the terminal equipped with surveillance photos of the travelers, but none of them could be found, the sources said.
|
# ? Feb 21, 2017 02:29 |
|
lolé
|
# ? Feb 21, 2017 02:37 |
|
tsayyy lmao
|
# ? Feb 21, 2017 02:38 |
|
LeftistMuslimObama posted:seems more like code injection op stole this for the subject of an internal email
|
# ? Feb 21, 2017 02:55 |
|
when you gaze behind the security theater curtain. in a better world this would lead to a purging of public/private partnerships and corruption charges. in our currently the gayest of worlds, prepare for full cavity searches being standard for "enhanced" interrogation of domestic travelers.
|
# ? Feb 21, 2017 02:57 |
|
rjmccall posted:stole this for the subject of an internal email im so honored
|
# ? Feb 21, 2017 03:01 |
|
I've got an Internet of poo poo fuckup in the making. I was asked to look at their proposed encryption flow for a device ~~with a chip in it~~ to talk to cellphones. It was as good as you can imagine. AES is good, right? * random key using random(). On embedded. You might as well just make it a compile-time constant. * encrypt using a ~~~secret~~~ fixed key, the same in every unit shipped. * send that to the other side. * encrypt all traffic using that "shared secret'. * ... with AES. Just AES, so I'm going to go out on a limb and guess ECB. I'm explaining to them exactly how loving awful that is, why they want a proper key exchange protocol and why they need a real AEAD instead of just saying "protected by AES". Unfortunately, I don't normally roll-my-own, I use vetted poo poo from experts so I'd want to double-check myself when putting the pieces together. My suggested improvements: Analog noise source feeding to an ADC, put that output into something like arc4random and use THAT for the keys. Implement proper KEX using ECDHE or something suited for embedded. Use an AEAD construct properly instead of AES in ECB. (EAX, maybe, to re-use the AES hardware block on the chip) Use crypto implementations vetted time-based side-channel attacks. I just know I'm going to become "the crypto expert" on this project, smdh. I'm "expert" in that I know some of the worst poo poo not to do. How hosed is my suggestion/am I?
|
# ? Feb 21, 2017 03:42 |
|
what's the mcu?
|
# ? Feb 21, 2017 03:52 |
|
hobbesmaster posted:what's the mcu? Was a bigger xmega, may be going to a STM32 for this. (And because Microchip bought Atmel and the prices all doubled overnight) E: without getting into identifiable information, it's a Thing That Opens With A Keypad that is getting turned into a TTOWAK-Or-Phone. Harik fucked around with this message at 04:17 on Feb 21, 2017 |
# ? Feb 21, 2017 04:15 |
|
stm32? its easier to do it right
|
# ? Feb 21, 2017 04:22 |
|
lol, an iot safe? loving amazing
|
# ? Feb 21, 2017 04:40 |
|
infernal machines posted:lol, an iot safe? More like unsafe.
|
# ? Feb 21, 2017 04:41 |
|
haw haw i bet teh secfuck has a smaller secfuck inside guys
|
# ? Feb 21, 2017 04:55 |
|
It's got CHACHA20-POLY1305 for AEAD but everything else is piece-parts. I don't trust them to put the piece-parts together correctly, they weren't even generating keys properly. So there's no _just_do_it_right() call that handles RNG seeding, KEX and AEAD for them. My original suggestion is only slightly modified: Entropy source (HW if they have it, otherwise sample a noise source a few thousand times and properly key expand the 10-bit values) use the STM32 crypto library for key generation use the library for KEX (I like EC25519 but we'll profile to see which is fastest) use CHACHA20-POLY1305 AEAD instead of naive AES ECB. Any remaining footguns?
|
# ? Feb 21, 2017 05:04 |
|
infernal machines posted:lol, an iot safe? It's funny now until it joins a botnet and DDOSs the forums. flakeloaf posted:haw haw i bet teh secfuck has a smaller secfuck inside guys You have no loving idea how right you are. I'm under no illusions this won't be a catastrofuck, the command protocol I shot down was full of direct unauthenticated-to-priveleged fuckups because they trusted the ~~app~~ instead of assuming it was hostile. I'm just trying to turn a 5-acre tire fire into a dumpster fire.
|
# ? Feb 21, 2017 05:07 |
|
Harik posted:It's got CHACHA20-POLY1305 for AEAD but everything else is piece-parts. if you want more foolproof mbed tls has you covered
|
# ? Feb 21, 2017 05:21 |
|
the only problem is...quote:mbed TLS (formerly known as PolarSSL) makes it trivially easy for developers to include cryptographic and SSL/TLS capabilities in their (embedded) products, facilitating this functionality with a minimal coding footprint. sounds like a challenge. someone is going to push keys of straight 0s into production
|
# ? Feb 21, 2017 05:23 |
|
yessssss
|
# ? Feb 21, 2017 06:00 |
|
LeftistMuslimObama posted:seems more like code injection op also this.
|
# ? Feb 21, 2017 06:01 |
|
Jesus you guys, nobody's pointing out the glaring MitM attack I left open? That's a pretty big secfuck. Added a signature to the KEX.
|
# ? Feb 21, 2017 06:03 |
|
i want to know more
|
# ? Feb 21, 2017 06:09 |
|
Word On The Street is that some projects pick up every hitchhiker they see/have their CI run on every commit
Jimmy Carter fucked around with this message at 07:19 on Feb 21, 2017 |
# ? Feb 21, 2017 07:15 |
|
uncurable mlady posted:tsayyy lmao flakeloaf posted:haw haw i bet teh secfuck has a smaller secfuck inside guys
|
# ? Feb 21, 2017 07:54 |
|
Jimmy Carter posted:Word On The Street is that some projects pick up every hitchhiker they see/have their CI run on every commit i thought that's what this bot literally is exploiting though, that that's what a lot of projects do as part of their automatic "is this PR valid" check?
|
# ? Feb 21, 2017 08:23 |
|
ate all the Oreos posted:i thought that's what this bot literally is exploiting though, that that's what a lot of projects do as part of their automatic "is this PR valid" check? yeah many projects are set up to use a service like TravisCI to auto-build PRs without any developer interaction. some jerk projects like swift think running arbitrary code without any review isn't a great idea, so you have to wait around for a committer to come by and push a button. tradeoffs
|
# ? Feb 21, 2017 08:35 |
|
ultramiraculous posted:yeah many projects are set up to use a service like TravisCI to auto-build PRs without any developer interaction. some jerk projects like swift think running arbitrary code without any review isn't a great idea, so you have to wait around for a committer to come by and push a button. For the most part you can just whitelist the guys who send you PRs all the time, so you only need to manually approve random yokels. That also makes a distinction between "mostly trusted" and "hey, we should check this guy out first" to avoid the temptation to just mash approve and grab a coffee to see which waiting PRs are going to be trivially rejected. I'm just waiting for some dumb node thing to get replaced with a bitcoin miner and have it deployed on every netflix server automatically, since webdevs have still failed to learn that lesson.
|
# ? Feb 21, 2017 09:12 |
|
Harik posted:I've got an Internet of poo poo fuckup in the making. lmbo
|
# ? Feb 21, 2017 15:24 |
|
some unironically great posts on this page nice work everyone
|
# ? Feb 21, 2017 15:25 |
|
cis autodrag posted:seems more like code injection op Thanks for whoever started this whole discussion. Been looking at CI/DevOps/buzzwordbingopipeline stuff at work and this has sparked some conversation.
|
# ? Feb 21, 2017 17:34 |
|
invision posted:Thanks for whoever started this whole discussion. Been looking at CI/DevOps/buzzwordbingopipeline stuff at work and this has sparked some conversation. keep in mind that ci is actually good if it is not configured stupidly. you just have to do different things when it's a public repo vs a private one and this bitcoin thing was targetting ci setups that were treating github as a trustable source of commits. if all your version control is internal it's actually pretty valuable for forcing dumbass devs to make sure their code actually compiles before they commit it.
|
# ? Feb 21, 2017 17:47 |
|
theres no reason to wait until it compiles to commit if you have a reason to commit. the CI system will only spit out compiled artifacts so it doesn't matter if a build breaks cause you should be using the last successfully built artifact instead of the source for the artifact.
|
# ? Feb 21, 2017 17:50 |
|
Shaggar posted:theres no reason to wait until it compiles to commit if you have a reason to commit. the CI system will only spit out compiled artifacts so it doesn't matter if a build breaks cause you should be using the last successfully built artifact instead of the source for the artifact. you would be shocked how often people will commit code that completely prevents a whole project from building. if your ci works right they will get an angry email from the ci server when the build fails.
|
# ? Feb 21, 2017 17:53 |
|
It does actually matter if a build breaks, because then you can't build.
|
# ? Feb 21, 2017 17:57 |
|
quote:Dear OSI bean dip, i only need an account to do online grocery orders that said the rumour is that they're unsure how they got breached even though it seems to be based on previous breach data (ie: from Ashley Madison or whatever) being used to get access
|
# ? Feb 21, 2017 18:01 |
|
Doom Mathematic posted:It does actually matter if a build breaks, because then you can't build. You can always build in the shaggar zone.
|
# ? Feb 21, 2017 18:02 |
|
|
# ? May 18, 2024 18:54 |
|
Shaggar posted:theres no reason to wait until it compiles to commit if you have a reason to commit. the CI system will only spit out compiled artifacts so it doesn't matter if a build breaks cause you should be using the last successfully built artifact instead of the source for the artifact.
|
# ? Feb 21, 2017 18:08 |