Some additional questions have been raised regarding Windows 8 SAS support, and a later deleted by owner separate question was posted about it, too. Since I've already started writing my answer to that question, and Windows 8 has also been mentioned in this thread, I'm thus posting it here. If th...
You think it's OK there? The question I wanted to answer was deleted by OP so.... ??
Seems like a script error, there is indeed a photo of it as an icon to some website under see also section. Looks like the script was written to select a photo based on wrong data, looks like it collects them in alphabetical order then displays the first one... strange.
@D3C4FF oh the usual bore for me I'm afraid ... been keeping busy, and now I need to call my agent to see if there's some more stuff for me that will keep me busy in the near future too ... getting rather tricky lately in this gowd forsaken land tbh :)
One of the things I need to do from time to time is to find subdomains of a site for example.
Starting with example.com
sub1.example.com
other.example.com
another.example.com
I'm looking for any additional ways to perform recon on these targets.
I'm currently doing a number ...
I find BGP Toolkit from Hurricane Electrics Internet Services quite useful for providing insight into any CIDR range, registrants, Who-Is info, IP and DNS info, IRR, assigned IPv4 and IPv6 CIDR ranges for individual ASNs, you name it. All data presented is also nicely linked to each-other for our...
Try with HE maybe you get lucky ;)
@D3C4FF You have any example domain we could try finding subdomains of?
I mean any that don't advertise some subdomains that you know exist, and we see if we can find them any other way?
@D3C4FF most would probably be in the range of say 3-8 letters and the charset is rather limited anyway... couldn't you create a script to test these in nmap?
nah, they have stupid things like sit-internet13.example.com
there are other ones like sft which i guess would be easy. I could probably 'brute-force' enumerate a list of short subdomains, then resolve them and if the IP's match my target IP's i'd be okay...
I expect you're searching Google with the site:*.example.com –inurl:www already? (this is too obvious but had to ask LOL) Have you cross-analyzed rDNS and FCrDNS results for discrepancies, maybe is something there like not updated DNS pointers?
oh and bloody robots.txt ... you forgot to list this in your question LOL
@RoryMcCune I think it was after the photo you posted, he just got stuck in my head. we were at a con, and I didn't know anyone, so I went and spoke to him, that was it really.
@LucasKauffman the problem is I generally feel sorry for the poor punter in the call centre, it's not them I'd like to shout at it's their management...
@RoryMcCune Tell me about it. I keep getting spammed by Barclays Bank asking me to take out a credit card with them. It seems there's a train of thought that says "inundating somebody with sales queries will eventually make them buy"
@ManishEarth be thankful you don't know what it is. it stands for Payment Protection Insurance, basically something you get A LOT of spam about in the UK
Anyone here know the steps required for starting your own CA? I suspect the question won't get a lot of love on ServerFault. (It's not a great question but it's a noble goal.)
I'm working on building PKI. I created RootCA, subCA with openssl. I know to know a steps to sign and using certificate's chain using openssl in Unix/Linux.
@Ladadadada The often neglected, but crucial step, is: understand what you are doing.
6
Usually I recommend using EJBCA, which offers graphical interfaces and documentation. Using OpenSSL directly is possible, but it requires some in-depth knowledge of X.509 (at least if you want to reliably make something secure).
@ThomasPornin I'm still making up my mind about this, but I think that even badly implemented SSL and/or PKI is an improvement over not having any in that it gets people used to expecting SSL as the default.
The counter argument is that it's a false sense of security which makes people take extra risks.
@Ladadadada There's actually documentation on doing this in OpenSSL itself, although, as @ThomasPornin says you need to know quite a bit to do it. For example, you won't be able to issue EV certificates at all without recompiling browsers, as these certs have OIDs built in, which are also compiled into the browser. I got as far as issuing certificates given a CSR, but there's also OCSP/CRLs to contend with, which gets complicated.
Having said that, they might have taken it out. I believe they were actually trying to discourage people from doing it.
@ScottPack actually, on a raised keyboard like the one pictured there the fail rate was about one in every 5-6, you can't quite make it out in the photo, but my fingers are contorted in an unusual way to avoid that
CPU penalty? Disk latency will be an issue long before CPU usage becomes a problem. And fixing HostNameLookups has nothing to do with any bottlenecks arising from large numbers of log files. Using SSDs for logging is a very silly idea. No, writing 250 log files is unlikely to be a problem - but NOT FOR THE REASONS STATED HERE — symcbean59 mins ago
When we say "pre-hash" we really mean "apply a hash function"
i.e. SHA-256(your-long-input), encoded in Base64 -> the input for bcrypt
If you go multi-bcrypt you have to make some assumptions on how bcrypt behaves has a generic hash function (danger !) and to fiddle with many salts and iteration counts.
@Tinned_Tuna If the hash function is resistant to preimages, the only remaining issue is competition for CPU: if the hash function used as first step is too slow, then it may force you to lower the iteration count for bcrypt. But that's not the case with SHA-256 (by a laaaaarge margin).
@Tinned_Tuna Developers should not, in general, be left to decide about cryptographic matters.
Right now, the best is to consider that bcrypt has a limited input length, and to live with it, by refraining from using passwords which are beyond that length.
This is a known defect of bcrypt, which should be corrected by designing a new, better construction, using the normal way these things are done: publication by cryptographers, review, more review, and even more review from other cryptographers, until they conclude that the new construction "seems good".
@Tinned_Tuna Actually the "community" thinks that scrypt is a good idea but not optimal, so there is an open competition for better proposals: password-hashing.net
Such functions are meant not only to be configurably slow, but also to be so that the best platform to compute them is exactly what the defender will use (i.e. a PC), so that the attacker will not get a boost by using specialized hardware (GPU, FPGA...)
apparently, scrypt still allows some speed boost through non-PC hardware
@lynks It is hard to integrate DRAM in an ASIC (IBM does it for the Cell CPU, but apparently it raises costs). ASIC/FPGA prefer SRAM -- and indeed modern FPGA have SRAM blocks of a few kilobytes.
@lynks Yep. DRAM is mostly one transistor and one capacitor per bit, while SRAM is six transistors per bit. Also, while DRAM requires refreshing, it draws less power per bit than SRAM.
@lynks When the RAM is not in the same chip as the CPU, then the latency advantage of SRAM tends to be lower, because the bus will include its own latency.
You won't get sub-nanosecond latency out of externalized RAM. To get the full potential of SRAM, you need to put in directly in the CPU
and that's expensive, because it increases die area.
1 megabyte of SRAM is about 50 millions of transistors
In existing CPU, you can have such amounts of SRAM, but clocked at a lower frequency than the full CPU, and semi-externalized (the CPU is actually several chips bundled together).
if bcrypt only takes 4k of SRAM, then an FPGA with 1MB of SRAM would be able to run (assuming optimal implementation, etc. etc.) 256 bcrypt processes in parallel
@lynks Oh, not that simple. You don't get one block; you get a lot of small SRAM blocks spread throughout the FPGA surface; you then hook these blocks with your custom circuits.
Which is good for the attack anyway: to make many parallel implementations, you need to be able to make many RAM accesses simultaneously.
@lynks When you compile your FPGA design, the placement tool will try to do that automatically, but you often have to help it along the way (it's a NP-hard problem after all).
@Tinned_Tuna A high-end FPGA will typically run at a few hundred MHz, say 200 to 500 MHz. Thus, each unit will be about 5 to 10 times slower than a CPU core. So it really depends on how many units you can put in your FPGA. It is hard to estimate "by hand" without spending some time on it.
@ThomasPornin I did a little FPGA work at university; I have to say I found HDL's to be a real brain twister. I was much more comfortable at highschool drawing out PCB designs and soldering in PICs :)
@ThomasPornin wait, I was under the impression that with an FPGA you were not creating a 'core' as such - in that it does not iterate over a set of instructions, you were actually creating the physical logic gates to perform the computation. So iteration speeds were huge, limited only by gate-delays.
@Polynomial fair enough, i always throught those 'cardlike' connections were fiddly, im a big fan of the big old DIN rings. you knew when you'd plugged your keyboard in back in the old days.
when nanoseconds count you need your lines to be perfectly equal in terms of resistivity and length
that's why you get those "squiggly" traces on motherboards - they're actually making the electrons travel further to make sure that the data edges and clock edges all align nicely in phase
@lynks I just entered 2nd page now that you mention it. Still quite some to go for 1st page but I consider it a success for 3 months here and not a single rep train answer (but that's my fault because I leave them for @Adnan and @Poly :P hehe). I also voted over 1100 times and only 25 are marked as down-votes (others got deleted). I see some forget to vote for whatever reasons...
@lynks +1 on that encoding question but now another question... if the question gets deleted eventually for being too localized, do you get to keep your rep points?
@TildalWave If the question is deleted, the rep points are lost (but not possible badges obtained with these points meanwhile). If the question is merely closed, the rep points are not lost.
I guess the lost rep is outweighed by gaining rep on casual up-votes on older answers in the mean time and maybe a bit more of that because of the higher total rep (yes, I've noticed LOL), but I'm still of an opinion that rep should stay (positive or negative change, doesn't matter).
@ThomasPornin oh you see this is what I was missing from the picture, thanks for explaining :)
@Iszi As per your request on meta, I've added the Linux perspective to that CTRL+ALT+DEL question. Hope it answers your question / satisfies your request. ;) (was quite some reading/writing involved LOL)
@Iszi I was also hoping I'll be less confused regarding the rationale behind it, but the more I was digging, it just became a bigger mess. I tried my best to reassemble my findings into a meaningful answer tho. The differences are clear IMO, but the why part - that isn't.
Maybe some of it comes from the general hatred towards what Microsoft stood for when Linux was first written (and also later)?
@Iszi I'm too old to hate, but no I wouldn't say all others are as old not to ;) Some even feel their support for this or that brand is non-trivial. It's been a while I was that young. :)
@TildalWave Yeah, i've read about multiple instances. Some 'researchers' publish the details on how to pwn them sometimes and let everyone free for all at em :P
@TildalWave No actually, forgot about watching that. I'll start listening to it now :P
One of the things I need to do from time to time is to find subdomains of a site for example.
Starting with example.com
sub1.example.com
other.example.com
another.example.com
I'm looking for any additional ways to perform recon on these targets.
I'm currently doing a number ...
I'm working on building PKI. I created RootCA, subCA with openssl. I know to know a steps to sign and using certificate's chain using openssl in Unix/Linux.
Please remove this question, I've figured it out.
