cybergibbons

I've seen enough badly implemented update systems (not on alarms) that really can't handle an update that doesn't go quite right. That's the advantage of being onsite - often the update is done using an ISP header which allows you to recover even if things go wrong. The same is true for BIOS updates on a lot of PCs. Some PCs have dual BIOS flash which means if one doesn't work after an update, it can fallback to a backup.

Lots of reasons: 1. It's an opening that wasn't there before to the heart of the system. 2. Quite a lot of systems don't do any checksum or validation of the firmware 3. Firmware updates can and do fail - leaving you with a system that doesn't work.

I'd much rather have a system where I can update the firmware than one I can't, and as part of that, the manufacturer needs to be open to dealing with reports of problems. Online firmware updates are a bit scary. There's a lot that can go wrong, and it's a big opening for attacks.

The biggest problem by far is the small key compared to the code size, so I've kind of stopped looking into this one.

The same is true for encryption in some wireless systems. The receiver in the panel doesn't have the overhead to deal with a key per detector, so it just uses a single system wide key. That means all detectors, ever, use the same key...

Thanks. With a) it seems that the main issue is that customers could reset the panel when the ARC doesn't want them to. b) I suspect exactly the same algorithm is used in the panel as in the program. I think the algorithm was written for whatever 8-bit processor was used in those days. Might as well show the algorithm used, the only bit of it that needs to be secrete is the vector at the top which I have changed. I don't know if you know any programming or python, but it's really simple. # Taken from the data in the exe vector = [3,3,5,5,8,1,3,9,8,0,5,9,3,9,4,1,1,0,9,4,3,0,2,2,8,4,3,2,8,4,9,4,1,3,3,3,3,8,5,3,0,2,4,3,2,1,8,9,0,5,4,3,9,5,8,3,9,9,1,0,0,9,9,3,3,8,2, 1,4,9,1,4,9,2,9,0,9,5,3,9,5,3,3,5,9,1,0,2,9,3,2,1,2,9,8,0,4,9,4,2,3,9,4,0,1,8,5,3,3,9,9,1,0,5,9,3,8,9,4,8,4,2,3,1,0,3,9,4,8,2,0,4,3,3, 1,0,5,2,8,3,3,5,2,8,3,2,9,5,2,1,2,4,4,3,0,4,2,3,4,1,8,2,9,1,0,5,1,8,2,4,3,5,1,0,3,8,5,3,2,1,1,3,9,3,2,3,5,8,3,9,0,3,2,3,5,3,8,4,0,3,9, 1,9,3,0,2,9,3,8,1,4,2,8,4,0,1,9,1,0,1,2,1,3,5,3,3,9,0,2,1,4,1,2,3,4,3,4,9,5,3,5,9,3,1,3,9,4,0,3,2,3,3,4,4,2,1] def generate_reset(quote, version): i = 0 tens = 0 reset = [] while i <= 4 : j = 0 result = 0 while j <= 4: offset = (version + tens + quote[j]) % 256 result = result + vector[offset] tens = tens + 10 j = j + 1 reset.append(result%10) i = i + 1 return reset print(generate_reset(quote = [0,0,0,0,2], version=131)) Notice that there is no multiplication, division, or anything fancy. The % symbol means "modulus" which most people know as "remainder". So "% 256" means "divide by 256 and give me the remainder". You normally get this for free in a microcontroller - an 8-bit number is limited to 256 values, so it just wraps round anyway. Panel firmware would contain it - in fact, now that I know certain panels have it, I can find the long string of characters called "vector" in some of them. The problem with panel firmware is that it is very hard for me to work out what is data and what is code. In x86 exes, there are normally a lot more hints available to me. I can also easily run an x86 exe and step through the running code to see how it works. With a microcontroller in an alarm, I can't easily do this.

Tunstall are someone who Menvier took over who Cooper took over? If you can point me towards a program to generate them...

Precisely Joe. This is just an easy target to show that.

It's all part of learning. Nothing lost for a few hours work.

So I wonder what drove the standard to require that and where the 5-digit Technistore code fits in to this? There must have been some reason behind it being 5-digit - it's rarely seen as a code length. It's a combination of the long code and short key. The key allows 256 distinct mappings between the quote code and reset code. That means that 12345 quote can only map to at most 256 of the 100,000 possible output values. 999,744 of the outputs are not possible - our keyspace has been reduced hugely. Notice I say "at most 256". It is possible for 12345 to map to 98765 using one or more keys. In fact, 12345 could map to 98765 using all 256 keys, but then we wouldn't need to find out they key at all. So if you tell me the reset code and I know the quote code, it is highly likely that I can just guess the key. For a very limited number of quote/reset pairs, I get 2 possible keys (in fact, there are two combinations with 4). So more than 99% of the time, I just need a single quote/reset pair to work out the key. So normally I get something like: 12345/74643 - only possible key is 123 (99.25% of the time) Sometimes I get this: 23654/34234 - two possible keys 232 and 154 (about .75% of the time) 98747/37265 - one possible key 232 (about 99.25% of the time) It would be really unlikely to get this: 23654/34234 - two possible keys 232 and 154 (about .75% of the time) 91737/72764 - two possible keys 078 and 154 (about .75% of the time) (we know the key is 154 as it is the only common one) Vanishly small chance of this happening: 23654/34234 - two possible keys 232 and 154 (about .75% of the time) 73748/38377 - two possible keys 232 and 154 (about .75% of the time) 98747/37265 - one possible key 232 (about 99.25% of the time) I've just tested these by running every single possible combination of key and input code against the algorithm. Not in any detail - have started looking at Texecom. Not aware of Tunstall.

Thanks Matt. You've raised something interesting there. With the 00-99 mapping system, you'd need to get all 100 pairs to be sure of the mapping. With Technistore, you need one pair to know the mapping.

My point is that someone, for some reason, decided a simple scheme of mapping 00-99 wasn't adequate. A much more involved scheme was developed, Technistore. This has the illusion of being more complex/secure, but once the algorithm is known, it is equivalent to a 000-255 mapping. The key is easy to derive from a single quote/reset code, and once the key is known, that's it. It is easy to develop a 00000-99999 mapping that uses a decent key length (128 bit is ideal, even 16 bit is much better) that would get rid of these problems. It wouldn't have required any more effort. It would have looked complex/secure, but also been secure, unlike with Technistore. Installers and ARCs don't seem to like the idea of customers resetting anti-codes themselves, so there has to be a security aspect here. Is it really a problem how bad Technistore is? No, not really. But what does it show? 1. Things can give an impression of being better, but they aren't really. 2. End users don't really have any way of knowing if things are better or not as they don't have the tools, knowledge or skill. 3. Some people developing alarms seem happy with this being the status quo. If I can write "AES-128" on my box, that's all they care about.

The way I see it, there are several stages this system has gone through: 1. Customers were allowed to reset alarms themselves. ARCs didn't want them to do this - what's the reasoning here? 2. A simple 00-99 quote and reset code system was developed and used by several panels. Someone must have deemed this inadequate because a more complex system was developed. Why was this required? 3. Technistore was developed which claims "military grade encryption" is used, and looks like it is licensed out to alarm manufacturers. So, the complexity of it has moved forwards incrementally over time. What was the reasoning for moving forwards? Were people finding that the 00-99 codes were being bypassed by customers? Or, in reality, was it just Technistore looking to make some money by artificially creating a need for security round this process?

How long is the anticode though?

How does the UDL software authenticate with the panels?

It is almost always without exception a bad idea to "roll your own" encryption: http://security.stackexchange.com/questions/18197/why-shouldnt-we-roll-our-own It's also a really bad idea to keep encryption schemes secret - the security should lie in the key, not the algorithm. If you keep it secret, the most clever person to look at it is going to be you. Make it public, and there is almost always someone more clever than you to take a look. I partly understand why Technistore is like this - it was implemented for embedded systems 25 years ago. Even with that in mind, it's got issues.

One of the big reasons I am here is to make sure anything I infer isn't false.

I don't recall any point where I haven't given enough evidence to back up a claim about a specific product. If I haven't named the product, it is because the manufacturer has made it clear they would be interested in legal action, so I need to be careful. The system that I didn't name that I don't think is good, I provided a document describing a similar system, and asked you to make your own conclusions. Open up a Scantronic wireless panel, look at that document, compare the radio modules, make your own judgement. I don't know. I'd question the use of the word "encryption" under trading standards. If your signalling system claimed it was encrypted and it turned out to be as weak as this, would that not be of concern?

Technically it's not encryption either. So, on a marketing and technical level, it's pretty bad. Where's the line? "This alarm uses rolling code" and the rolling code is 1,2,3,4. Is that dodgy? So if I am allowed chosen plaintext (i.e. I can call up the ARC and tell them whatever quote code I chose, and get a response), then it wouldn't require many pairs to get the keys. I don't know how possible this would be, as I think they would have to see an alarm activation, which means I would need a real quote/code pair. If it's only known plaintext (i.e. I am using valid quote codes generated by the alarm), it would be quite a lot more pairs required. Still a tiny number compared to the security a 2048-bit key affords. All of this would have been caught by an undergraduate doing a cryptography coursework "Is this homebrewed MAC secure?". It wouldn't have been hard to make this secure at all. Actually, I think it would be less effort just using something ready made.

How do I change that though? I've looked at a good few systems, enough that I can form an opinion of where they lie in terms of security. I've posted information on why I think the bad products are bad, some of which has been in quite a lot of depth. I can go into more depth, but as many have said, it would be beyond them. I can that installers aren't in possession of all the facts - there are alarm systems that fall far short of the marketing.

Absolutely nothing, it's marketing. It suggests it would be a standard that the military could use, which suggests it might pass some standards that the military have. If any of you have Technistore in front of you and it is a version where you can change the seed, try this: Seed 100, code 33333 Seed 101, code 22222 Seed 102, code 11111 Notice how they all produce the same unlock code? It's leaking information - changing outputs in a predictable way like this shouldn't produce a predictable output. I think, but I am not 100% sure, than it would only take about 50 valid reset/code pairs for me to determine the seed and the far longer initialisation vector (256 bytes). So even if the key was much longer, the algorithm sucks.

I don't think I have said that one brand is secure really, just my impression of it is better than others. Is it any different to an installer saying they prefer Texecom over Honeywell?

For Technistore, on average you need just one quote/reset code pair to derive the seed code. About 0.25% of code pairs lead to two valid seeds, and less than 0.01% generate more than that. So after a single reset, you have the seed for your panel, and it seems quite likely the seed for all alarms on the same ARC (correct me if wrong, there are quite a lot of references to the seed not varying on a per-customer basis). Like I say, the key length is so short that you can normally recover it with a single quote/reset code pair. No need to spy on the installer. So what if it isn't genuine? The point is that this mechanism is touted as secure ("a military strength data encryption algorithm") and it isn't. There isn't a need for it to be insecure, this is just bad code. Reverse engineering for the purposes of writing your own code for interoperability is specifically protected in law in the EU.