The Market Ticker
Commentary on The Capital Markets- Category [Technology]
Logging in or registering will improve your experience here
Main Navigation
MUST-READ Selection(s):
Don't Do It Lennar -- Talk To Me Instead

Display list of topics

Sarah's Resources You Should See
Sarah's Blog Buy Sarah's Pictures
Full-Text Search & Archives
Legal Disclaimer

The content on this site is provided without any warranty, express or implied. All opinions expressed on this site are those of the author and may contain errors or omissions.


The author may have a position in any company or security mentioned herein. Actions you undertake as a consequence of any analysis, opinion or advertisement on this site are your sole responsibility.

Market charts, when present, used with permission of TD Ameritrade/ThinkOrSwim Inc. Neither TD Ameritrade or ThinkOrSwim have reviewed, approved or disapproved any content herein.

The Market Ticker content may be sent unmodified to lawmakers via print or electronic means or excerpted online for non-commercial purposes provided full attribution is given and the original article source is linked to. Please contact Karl Denninger for reprint permission in other media, to republish full articles, or for any commercial use (which includes any site where advertising is displayed.)

Submissions or tips on matters of economic or political interest may be sent "over the transom" to The Editor at any time. To be considered for publication your submission must include full and correct contact information and be related to an economic or political matter of the day. All submissions become the property of The Market Ticker.

Considering sending spam? Read this first.

2018-05-24 13:50 by Karl Denninger
in Technology , 179 references
[Comments enabled]  

I told you so.

A Portland family contacted Amazon to investigate after they say a private conversation in their home was recorded by Amazon's Alexa -- the voice-controlled smart speaker -- and that the recorded audio was sent to the phone of a random person in Seattle, who was in the family’s contact list.

Or, for the reading-impaired:


Look folks, I get it.  Home control and monitoring is both cool and powerful.

And useful.

Putting that data, ever, in the "cloud" is not only a supremely bad idea even if everything works 100% bug-free because the designers could intend to screw you with the information the facts are that there is no such thing as bug-free software, ever, period.

I get it.  You want the capability.  Good.  You can have it -- without giving up any vestige of privacy in your home.

I've got the package to do exactly that; does anyone want to buy, own, distribute and make a ton of money selling it?


1. It never talks to the cloud.

2. It does not store authentication data (e.g. login and password) anywhere off your premise, including on your phone.  It gets a token with a validity time you choose on the phone (or browser, if you're talking to it over a browser) but never saves the authentication data anywhere except where it must, on the gateway itself, and there it is stored hashed.

3. It can do real-time video and images secured as well from common and insecure commodity-priced IP cameras; essentially all current such video cameras stream using unencrypted data streams, which means not only can anyone pick them off companies like Amazon, if you let them have "cloud" storage access to said data, can and must be presumed to be using visual recognition technology (e.g. facial recognition, etc) on that data.

Oh, and it has a licensing and security model based on SSL using a private CA (which the buyer of course would own and be able to customize to suit) so there's no ongoing demand to buy a public certificate either.

I know someone's out there who wants to cut the crap in this regard.  No, it won't take voice commands because there is no way to do that without severely compromising your privacy.  Yes, you must tap your phone screen. And?  You do that now every day -- right?

The asking price isn't a billion dollars, such as what Amazon paid for "Ring" either -- it's much, much more reasonable.  In fact if you're running a real business (or intend to) it's definitely within reach.

Come and get it -- first come, only served.

View this entry with comments (opens new window)

2018-05-24 07:00 by Karl Denninger
in Technology , 124 references
[Comments enabled]  

Whadda 'ya mean a car in "self-driving" mode ramming the rear of a stopped truck isn't considered "acceptable performance"?

The travel group found that 73% of American drivers say they would be “too afraid” to ride in a fully autonomous vehicle, up from 63% in late 2017. Also, nearly two-thirds of U.S. adults would feel less safe sharing the road with a self-driving car while walking or riding a bicycle.

“Despite their potential to make our roads safer in the long run, consumers have high expectations for safety,” Greg Brannon, AAA’s director of automotive engineering and industry relations, said in a statement.

I have the potential to screw a Dallas Cowboys' cheerleader too.

Of course what's not disclosed is the odds of such potential working out for me.

Or for the AAA screamers.

Or for you, for that matter.

Oh, and if your car of choice is a Tesla it appears they come with a built-in crematorium pre-loaded with fuel for you too, if you happen to crash them the wrong way and can't get out of immediately (like, for instance, because you're unconscious!)

I don't suppose ramming another vehicle at highway speeds would be considered "operating as designed", and anyone who tries to tell me that a fire truck is not "reasonably visible" in front of the vehicle...... well, pull the other one eh?

Reality is that "out-of-scope" processing, otherwise known as sentience, is an inherent part of safe vehicle operation and no computer has ever demonstrated that.  So-called "AI" is nothing of the sort as no machine has ever demonstrated out-of-scope deduction.

Why?  Because both human and non-human hazards on the roads are non-deterministic, that's why.  To get rid of one (the human hazards) you'd have to prevent non-automated, non-100% communicating (that is, spying on you all the time) vehicles from the roads entirely which includes pedestrians, bicycles and motorcycles and to prevent non-human hazards you'd have to ban rocks, deer, bears and other things that can and do intrude on roadways - including things like "gators" (tire casings from trucks.)

Good luck with that.

View this entry with comments (opens new window)

2018-05-23 12:59 by Karl Denninger
in Technology , 84 references
[Comments enabled]  

Might want to read this Ticker again....

As of now this problem is, generally-speaking, solved.

Your IP camera does not need to be visible from the outside.  At all.

You also never need to store its password anywhere outside -- not even on your phone.

The HomeDaemon app also never stores the password to either the camera or the HomeDaemon controller; it authenticates using a key that it maintains while running (and can be set to run in the background), and the controller, on demand, negotiates with the camera, gets the unencrypted stream, encrypts it using SSL and a private-CA secured certificate (that is essentially unbreakable), and displays it.

There remain some issues with bandwidth consumption and for that reason I'm not currently using the highest resolution stream capacity available (especially on the 2k+ cameras!) but I have managed to get the latency down to roughly 1 second, which isn't bad at all.

All this on a Pi2, which has about a quarter of the power of the newer Pi3 series.

Encapsulating with the higher resolution and lower bandwidth consumption options is being worked on, as is the ability to move the camera both to presets and arbitrarily, along with setting the camera's presets.  Those latter capabilities are, since I have the camera interface worked out, simply a matter of adding the buttons to the screen.

Are you in the business of providing home automation solutions or selling houses with so-called "smart" features?  This is the one you want.

You can not only control and monitor everything in your house with real time notifications, zero cloud storage or use (therefore nothing to steal!) but the system also integrates fully with any of the Amcrest camera family.

Look to the right for contact info; it's available right here, right now.  Buy it all and make a fortune.

View this entry with comments (opens new window)

2018-05-18 09:55 by Karl Denninger
in Technology , 82 references
[Comments enabled]  

Ah, the idiots who write code are at it again.

This time it's pretty bad too.  Apparently nobody thought about the problem of malformed clients accessing HTML components in an email that happens to be encrypted.

That could be trouble.  You see, HTML can post things as well as read them, and along with specific sorts of HTML markup and similar very evil things can be done -- including sending the contents of an encrypted mail outbound.

The problem is that MIME allows for multiple parts, so if you can steal an encrypted message you can then send it as a piece of a new one, which the client will dutifully (as it has the key!) decrypt it for you and then send it out to the bad guy.

Ain't that nice?

It doesn't work if it's a text email, because there are no multi-part pieces in such a message.

Of course the obvious is don't allow HTML components and methods in an encrypted message, and if a sender tries that then don't act on them at all on the receiving end.  Gee, nobody thought of that one eh?  smiley

The Thunderbird folks already have figured out how to fix it (duh!) but the bigger problem is going to be Outlook, which has support for this (albeit broken in other ways, but people do use it) and which has several older incantations laying around that are probably well outside of Microsoft's willingness to support and update.  If you're using one of those you're probably screwed.

Exploiting this requires stealing some of the ciphered email you want to decode, so you have to break into the target's machine first.  But the entire point of PGP or S/MIME is to make stealing the recipient's computer worthless.


View this entry with comments (opens new window)

2018-05-17 10:08 by Karl Denninger
in Technology , 142 references
[Comments enabled]  

You're not going to like this.

Got a "Ring" doorbell?  Or, for that matter, any of those other nice IP cameras?  Doesn't matter who makes them, by the way.

Most security-conscious people are aware that there's a huge problem with allowing any sort of "cloud" storage option to be turned on.  Many people don't care, but you damn well should.

But here's the other problem that comes with these things: The video stream itself is totally insecure and your credentials are not much safer.

I'm unfortunately forced to recommend at this point that all such devices be immediately shut down in terms of off-premise access to the video.  Period.  Full-stop.  Right now.

The reason is a bit complex, so hopefully you'll read the whole thing to understand it.

Here's a typical start-up transaction for real-time video streaming to one of these cameras:

13:15:16.697677 IP (tos 0x0, ttl 63, id 20712, offset 0, flags [DF], proto TCP (6), length 163)
D2.Denninger.Net.50501 > Flags [P.], cksum 0xc419 (correct), seq 1:112, ack 1, win 343, options [nop,nop,TS val 7363585 ecr 37265593], length 111: RTSP, length: 111
OPTIONS rtsp:// RTSP/1.0
CSeq: 1
User-Agent: Live555

I want to look at the camera in real-time, main channel and normal "substream" (different resolutions, etc) using RTSP (streaming video) > D2.Denninger.Net.50501: Flags [P.], cksum 0x2c43 (correct), seq 1:143, ack 112, win 905, options [nop,nop,TS val 37265595 ecr 7363585],length 142: RTSP, length: 142
RTSP/1.0 401 Unauthorized
CSeq: 1
WWW-Authenticate: Digest realm="Login to AMC000PD39KR3820JT", nonce="da8

Heh jackass, you didn't give me any sort of credentials -- go away or tell me who you are with a password.

13:15:16.724851 IP (tos 0x0, ttl 63, id 20714, offset 0, flags [DF], proto TCP (6), length 394)
D2.Denninger.Net.50501 > Flags [P.], cksum 0xb072 (correct), seq 112:454, ack 143, win 347, options [nop,nop,TS val 7363588 ecr 37265595], length 342: RTSP, length: 342
OPTIONS rtsp:// RTSP/1.0
CSeq: 2
User-Agent: Live555
Authorization: Digest username="karl", realm="Login to AMC000PD39KR3820JT", nonce="da8684cec2ea70ff015538fb006139e3", uri="rtsp://", response="20dc9cda80205b5d2d8f2ae9c335dc27"

Ok, I'm Karl and here's a password for that previously-requested video stream.  Can I have it now?

Note that there is no actual password.  There's a "nonce" and a "response", both of which are not clear text.  The reason is that the camera (thankfully) demanded "digest" authentication.  Note that some earlier versions of camera software allowed "basic", which transmitted credentials -- including the password itself -- in plain text.

Amcrest shut that off about a year ago which is a good thing.  When they did they broke a lot of third-party software that, believe it or not, actually used and relied on plain-text passwords being sent over the wire.  That's so stupid it belies basic logic, but there was a lot of third-party code out there that did.  The fact that Amcrest forcibly disabled this by removing the option from their firmware is one of the reasons I've sort of liked their cameras -- they at least try.

The problem, however, will become clear momentarily....  let's continue.

13:15:16.741703 IP (tos 0x0, ttl 64, id 4151, offset 0, flags [DF], proto TCP (6), length 210) > D2.Denninger.Net.50501: Flags [P.], cksum 0x9460 (correct), seq 143:301, ack 454, win 972, options [nop,nop,TS val 37265597 ecr 7363588], length 158: RTSP, length: 158
RTSP/1.0 200 OK
CSeq: 2
Server: Rtsp Server/3.0

I'm the camera, I like you and your password is good.  Here is what I know how to do with the RTSP protocol.  Please tell me how to proceed.

D2.Denninger.Net.50501 > Flags [P.], cksum 0xa4a0 (correct), seq 454:822, ack 301, win 351, options [nop,nop,TS val 7363590 ecr 37265597], length 368: RTSP, length: 368
DESCRIBE rtsp:// RTSP/1.0
Accept: application/sdp
CSeq: 3
User-Agent: Live555
Authorization: Digest username="karl", realm="Login to AMC000PD39KR3820JT", nonce="da8684cec2ea70ff015538fb006139e3", uri="rtsp://", response="55f38faa28da02835fdfd2de248f1632"

Ok, please tell me what the stream that is identified as channel 1, subchannel 0 looks like.  Oh, and here's authentication credentials again (note the "response" is different, because the hash includes the command, in this case "DESCRIBE" (with parameters) although the nonce has not been re-generated (this is part of problem #3 I'll get to later)


13:15:16.759068 IP (tos 0x0, ttl 64, id 4153, offset 0, flags [DF], proto TCP (6), length 773) > D2.Denninger.Net.50501: Flags [P.], cksum 0xbee2 (correct), seq 301:1022, ack 822, win 1039, options [nop,nop,TS val 37265599 ecr 7363590], length 721: RTSP, length: 721
RTSP/1.0 200 OK
CSeq: 3
x-Accept-Dynamic-Rate: 1
Content-Base: rtsp://
Cache-Control: must-revalidate
Content-Length: 506
Content-Type: application/sdp

o=- 2252311096 2252311096 IN IP4
s=Media Server
c=IN IP4
t=0 0
m=video 0 RTP/AVP 96
a=rtpmap:96 H264/90000
a=fmtp:96 packetization-mode=1;profile-level-id=640029;sprop-parameter-s
m=audio 0 RTP/AVP 8
a=rtpmap:8 PCMA/16000

Here's a bunch of information for you describing the media you requested.  The frame rate, the format of it, the audio that's included and its bitrate, etc -- all of which you'll need in order to successfully decode and display the video.  Oh, and your password is still good because I said "ok".

D2.Denninger.Net.50501 > Flags [P.], cksum 0xbc55 (correct), seq 822:1249, ack 1022, win 357, options [nop,nop,TS val 7363592 ecr 37265599], length 427: RTSP, length: 427
SETUP rtsp:// RTSP/1.0
Transport: RTP/AVP/TCP;unicast;interleaved=0-1
x-Dynamic-Rate: 0
CSeq: 4
User-Agent: Live555
Authorization: Digest username="karl", realm="Login to AMC000PD39KR3820JT", nonce="da8684cec2ea70ff015538fb006139e3", uri="rtsp://", response="a069805debe9e14f99804d37242eac50"

Thank you.  Please set up to play the stream in question (I've decided I can understand it), and by the way, here's another hash (with my password of course) just so you know it's really me.

13:15:16.776718 IP (tos 0x0, ttl 64, id 4154, offset 0, flags [DF], proto TCP (6), length 195) > D2.Denninger.Net.50501: Flags [P.], cksum 0xf8ba (correct), seq 1022:1165, ack 1249, win 1106, options [nop,nop,TS val 37265601 ecr 7363592], length 143: RTSP, length: 143
RTSP/1.0 200 OK
CSeq: 4
Session: 372956013002;timeout=60
Transport: RTP/AVP/TCP;unicast;interleaved=0-1;ssrc=6D23767F
x-Dynamic-Rate: 1

Ok, you're good.  We're ready to go whenever you are, but the session I'm setting up for you is only valid for the next 60 seconds.  I'm going to send it over TCP, here's a session ID so you know it's the right one when it starts and oh, your password is still ok.

13:15:16.780403 IP (tos 0x0, ttl 63, id 20717, offset 0, flags [DF], proto TCP (6), length 435)
D2.Denninger.Net.50501 > Flags [P.], cksum 0xb842 (correct), seq 1249:1632, ack 1165, win 362, options [nop,nop,TS val 7363594 ecr 37265601], length 383: RTSP, length: 383
PLAY rtsp:// RTSP/
Range: npt=0.000-
CSeq: 5
User-Agent: Live555
Session: 372956013002
Authorization: Digest username="karl", realm="Login to AMC000PD39KR3820JT", nonce="da8684cec2ea70ff015538fb006139e3", uri="rtsp://", response="b215d803b422a92152c11d1abcf94387"

All good.  Start the stream please, play from time 0.000 on the session ID you previously set up for me.  Oh, and here's my credentials again.

13:15:16.816148 IP (tos 0x0, ttl 64, id 4156, offset 0, flags [DF], proto TCP (6), length 176) > D2.Denninger.Net.50501: Flags [P.], cksum 0x97a8 (correct), seq 1165:1289, ack 1632, win 1173, options [nop,nop,TS val 37265605 ecr 7363594], length 124: RTSP, length: 124
RTSP/1.0 200 OK
CSeq: 5
Session: 372956013002
Range: npt=0.000000-
RTP-Info: url=trackID=0;seq=32390;rtptime=2924735

Here it comes! > D2.Denninger.Net.50501: Flags [P.], cksum 0x4d5c (correct), seq 1289:1341, ack 1632, win 1173, options [nop,nop,TS val 37265609 ecr 7363601], length 52: RTSP
13:15:16.871343 IP (tos 0x0, ttl 63, id 20719, offset 0, flags [DF], proto TCP (6), length 52)
D2.Denninger.Net.50501 > Flags [.], cksum 0x9535 (correct), ack 1341, win 362, options [nop,nop,TS val 7363602 ecr 37265609], length 0
13:15:16.962682 IP (tos 0x0, ttl 64, id 4158, offset 0, flags [DF], proto TCP (6), length 1500) > D2.Denninger.Net.50501: Flags [.], cksum 0x5aa7 (correct), seq 1341:2789, ack 1632, win 1173, options [nop,nop,TS val 37265619 ecr 7363602], length 1448: RTSP
13:15:16.962846 IP (tos 0x0, ttl 64, id 4159, offset 0, flags [DF], proto TCP (6), length 1500) > D2.Denninger.Net.50501: Flags [.], cksum 0xdfe3 (correct), seq 2789:4237, ack 1632, win 1173, options [nop,nop,TS val 37265619 ecr 7363602], length 1448: RTSP

And there it is.... there's a lot more of course but these are the start of the packets containing the actual video.

Now here are the problems, in order:

1. RTSP is unencrypted. This means that the actual video is flowing over the network with absolutely zero encryption of any sort, and anyone in the middle can pick it off.  Watching it on a WiFi network that is unsecured in some coffee shop?  Everyone within 200' of you can see your video.  Is that your cute kid on the "baby monitor" or your empty house?  That wouldn't be a problem, right?

2. Every three-letter spook and criminal malefactor who can access any part of the infrastructure between you and the camera can also trivially decode and display that video in real-time.  Yes, I said anyone.  It requires exactly zero coding or effort to display an RTSP stream you capture; simply feed it to any media player that understands the format and voila -- there you are.  This means that any time you have an actual video stream running anyone who is sufficiently motivated can watch it too at the same time you do or grab the stream, save it to their device and watch it whenever they'd like.  There is not a single online store or financial institution that finds this sort of crap acceptable which is why they all have those little padlocks (https) on their web pages.

3. MD5 is not secure.  That's what "digest" uses to hash those passwords.  It's better than sending them in plain text over the Internet, but not that much better.  While someone who gets a single session would probably have trouble breaking your password someone who manages to get a bunch of these negotiations over time absolutely can do so.  I will note that MD5 was deprecated a long time ago as a sufficiently good digest for hashing passwords in general and it's also not considered acceptable as a hashing algorithm in SSL either, for the same reason -- it's not very hard to break.  If your camera is connected to any sort of "central" or "cloud" service those transactions are all going to one destination and thus you must assume your password has been stolen.  Once stolen, of course, said bad actor can now break into your camera at any time in the future, not just when you're watching it.  Again, I repeat: If you allow your "IP camera" to connect to or if you use any sort of "cloud" storage, control or similar facility you must assume that the login credentials have been compromised.

I've made a "command decision" when it comes to HomeDaemon's app -- it's not going to support doing that as it stands although literally every single damn app out there right now does.  No way am I doing that with my code as it's a blatant and severe security problem.

Instead what I support now is SSL-encrypted transport of captured stills, which is easy because I already use SSL-encrypted transport to talk to the server, the server has the ability to ask the camera for stills already, both are on the same local LAN which is encrypted with AES (and thus reasonably secure) and HomeDaemon-MCP (the server code) already knows how to enforce permissions for sessions so you not only need to authenticate but some users can have access to the camera images and others not, as you wish.  That is secure off-premise right here, right now because (1) the camera is not directly sending the picture, HomeDaemon-MCP is, (2) it demanded and got an SSL connection to the client on the phone and (3) the transport is thus safe from interception or even knowledge that you made the request.

So if you click a camera icon on the HomeDaemon-MCP app what you'll get is the most-recent still the camera has captured, usually configured on the base code to be snapped whenever the camera "sees" movement -- instead of a real-time video stream.

What I'm investigating supporting (if I can figure out how to feed Android's media decoder and display tools bidirectionally with arbitrary streams of data) is a secure, SSL-encrypted tunnel for that video generated by HomeDaemon-MCP and the app, removing the risk for video displayed through the app.  You can then absolutely block all outside access to the cameras, period, at your firewall without losing functionality.  The power consumption on the phone doing this might make you grimace due to the data rate involved but if I can figure it out the transport will be secure.  My first blush look at this is that it's not something either the stock mediaplayer or Exoplayer were designed to handle; the former is part of the framework and not designed to be modified at all, but the latter is an open-source project.  In any event it doesn't look like a "quick" thing to support, if it's even reasonable to do at all given the current state of the Android codebase.  It might not be.

I suspect the reason there is no "RTSPS" is that these cameras simply don't have the CPU horsepower to run SSL encryption for a video stream at the data rates required without puking, and fixing that would massively increase the cost of the cameras.  That makes sense but it means that every single IP camera out there right now is trivially intercepted by anyone sufficiently motivated to do so when in actual use to view video, and any connected to a cloud account gives anyone sufficiently interested a nasty and unencrypted "choke point" through which to collect enough digest information to break your password.

I repeat: As it stands right now Internet-accessible streaming cams at minimum expose the actual video in real time to anyone who cares to intercept it without any sort of real effort and any connected to a cloud service of any sort almost-certainly wind up divulging enough information to make compromise of your password and thus permanent access to same quite easy.

Psst.... if you're in the "home control and monitoring" business, or anything associated with it.... yes, the entire codebase is for sale as I've pointed out before.  Who wants to have the competitive advantage of their systems not being trivially watched by the neighborhood crook say much less organized rings of same?  Email me.....

View this entry with comments (opens new window)