-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 Tor2web ======= And for the Tor2web idea, what about cloaking? changing the output depending on who connects, if Tor users connect show the Tor hidden services, if normal users connect show them only the normal web, Google might also be able to do this filtering when it reads the X_Tor2Web header. Lets assume Tor users by using the Tor network have agreed they might see censored and possibly *bad* content now they cannot complain. But it occurs to me:- A) You would need a way to tell Tor user requests from normal user request and Tor intentionally prevents this (you could send a HTTP header with the request for example X-AllowHidden to access the proxy's extra functionality). B) If this was the case you wouldn't need it because Tor users would already be using Tor to access the internet anyway. C) The government would always use the X-AllowHidden request header to see censored content so it wouldn't help with that.
The Mindware Satalite Network ============================= I am currently working on the Mindware Satellite Network, I don't have a web site and very few people know of the existence of this project. Current weaknesses ================== It occurred to me currently the greatest weakness of Tor, I2P (and many others although very good designs) are vulnerable to end to end analysis, which ever way you cut it because of the way real time protocols work you can detect hosts by whether they are online or not. The UK government wants to log "Who talks to who, when and for how long". Usually when Alice and Bob communicate they can see *who* is communicating but because of encryption cannot read *what* is said, I reckon we can do better than this, I recon we can deny them who is talking to who aswell. What makes a good network ========================= o Encryption (without question) *hides what is sent*. o Protocal obfuscation, this actually proves to be quite difficult in practice and now it appears they are also copying our protocals to trick and probe our software aswell. *hides protocal /software used* o Connection redundancy, multiple hops to hide the real source. *hides who is sending and where data is going* o Traffic redundancy, cover data to hide when or even if actual data is sent. *hides when, size and how long or even if data is sent* o File/resource redundancy and caching, files stored in multiple places to make it more resistant to churn, attack and failure (note: DHTs are good, direct and scalable for storing stuff but it appears to be quite difficult to hide stuff in DHTs, although I will encrypt files with a key, im working to see if theres a solution to this). The ideal here is that you are able to "hide" files in the DHT. o Capacitance, ability of the network to store and time shift data. An often lacking property, which could potentially destroy end to end analysis. Like email systems store emails on a server until your online to collect them, in this case data is stored in the network itself, this disconnects the time the sender and reciever is online. With the hypothetical network you can send to a person and make requests *even* when the other person is offline unlike TCP/UDP which only works when the host is online, the network stores the messages, either you send the messages and they float in the network until the other person comes online, they float in transit (may avoid legal issues), or you upload files to the network, which the other person can later download. This is totally different to how normal instant TCP/UDP networks work and will require totally new and very careful software design. Files are no longer stored on hosts, which protects the uploading client. *careful time shifting and DHT use hides when a user/host is online*. Current designs =============== Tor provides Encryption, Protocal obfuscation, onion routing but no cover data. And because it uses the normal web and client/host paradigm has no capacitance, file redundancy and is susceptible to the client/host correlation attacks mentioned above. But it works great for what it was designed for as an encrypted tunnel. I2p has the same problems because it access's the client/host real time web but I2P does additionally have the ability to use cover data which is probably not implemented. Its probably only weakness is clients/host could be correlated by if they are online or not. Freenet does file redundancy and capacitance but only has weak side effect onion routing. It occurs to me if Freenet had a HTTP proxy, and a Tor exit was tunneled straight into Freenet it will solve all problems at once as long as the junction was not logged/monitored. Abstract ======== I thought what if you connect to 5 peers, they connect to 5 excetera, you transmit to all 5 once a second like a beacon, every packet is 2K encrypted, every send the packets changes, every hop the packet changes, it always looks different *even* if its the same data (I may also make an option to wrap it all up to look like HTTP requests so it goes through firewalls and DPI filters), in effect you are connected to *everyone* on the network at once, they can see you talking to 5 peers, but they have no idea where the packets are really going, traffic analysis is *impossible* (peer-peer comms can be monitored and peers discovered by they know nothing about the real data transmitted). Packets are encrypted peer to peer, between links, they are also encrypted end to end. I will need to do prioritization so communication data is sent first, and file transfers are less important and queued. This has several advantageous and disadvantages, low data transmission means it will be slow, it limits the data sent but it also limits the data lost. It might also be possible to find out where things are by how long (how many hops) it takes data to come back, but with all the other data transmitted this may not be a problem. The web is to much a client-server fixed location system (including all its old TCP protocals HTTP, email, etc) in my opinion instead of trying to send an encrypted tunnel through mud into mud, it needs to be re-written from the ground up as an efficient peer-peer (probably DHT) system. But at this time I cannot envision how Socks could be re-written to accommodate such a better system or what the addressing system would look like. Where as the normal internet is direct and real time more like a TV get it instantly, but is easy to track, the Mindware Satellite Network is more like a mail order catalog send a request now pick it up later. The Mindware Satellite Network will be more for communication than file sharing, but the DHT could be used for both. You will notice that my design does Encryption and cover traffic by design and Protocal obfuscation and weak onion routing as a side effect. I will also likely do onion routing by bouncing packets within the network also. Note: If we all used 2K random looking packets sent at a regular rate, we could all be obfuscated just by all looking the same. Only problem with this method is if they see 2K packets it would be a fair guess it was peer to peer traffic they where looking at. Completed work ============== I have completed my last project Lockbox2.0.0 Alpha which provides transport independant, node to node encryption, a little like PGP but uses symetric keys. and is now ready for testing. I have a bunch of code already for my Satalite system:- encryption, hashes, compression. And some modules that are nearly finished:- coolcomm, http module, http programs. Ive already coded Cool comm which I use as my Base layer, it provides TCP or UDP peer to peer communication, packet sequencing and automatic resending, attack protection (ignores everyone until they're authenticated), authentication, link encryption (does not have diffie yet or scalable authentication), IP and user blocking lists. I need to add in a few new connection types, my new 2K encrypted packet object, and I will make a connection type that first send a http header so it looks like a http stream. It can do either UDP or TCP so TCP can easily be turned off in case it turns out TCP is too vulnerable. Needs finishing. Not sure whether to have Cool comm running as a service, use DDE to communicate with it and have a DLL api so other programs can create and use comm connections directly, only problem is this might allow attackers to attack the system more directly so not sure if this is a good idea. Done some socks code because I will likely do a Socks interface input but it is not really designed for real time TCP/UDP. I have working but incomplete http module works but need finishing, I will need to finish it so I can finish my http client and server to run over it (the server and client have no script functionality to reduce attack vectors). I have working but incomplete http client and server. Current work ============ Im now working on designing and implementing the routing layer. I plan on making the routing system accessible so other programs can use the routing system. I would rather use a simple input like Socks for programs to use, so normal programs can be easily interfaced with it, however the system isn't really compatible end-end TCP, an onion address variation is possible, so is UDP socks or I may use DLL access instead (again not sure if this is a good idea due to attacks). I am doing a test Kademilia program to do DHT routing and "object" storage, I will probably store files, user identity's and domains names (as in web site start pages) in the final DHT structure. So you can request files and web pages and send to a users inbox at any time, no single person is responsible for the DHT storage. DHT routing will be a fast, efficient and a more or less direct way of getting packets to destinations and I wont need to know where the destination is to send, this is the faster method, but with this network structure it will be "open" requiring at least one seedpeer in a fixed location and *might* allow the adversary to easily find peers, this will be faster but may not be most ideal way for hiding network nodes. In progress. I also plan on being able to use it in darknet mode, only connecting to known trusted friends, I am currently working on how to do efficient packet routing for darknet connections. I will also need the algorithm to be adaptive if nodes are slow, or malicious nodes deliberate drop packets such as byzantine attacks. I am currently working on how this would be done so I can operate with both. It will likely be something like Kademelia, adaptive darknet routing, and when no other options available flooding. I thought packets once sent could float in the network until the destination comes online, packets could even be "intelligent" in that once sent they find there own way to the destination. Planned work ============ Once routing is coded upto a certain standard and I have an API I can write apps on top of it. Not sure if I will do DHT storage as a seperate app (maybe a piece by piece upload/downloader like bit torrent, I wrote one just need to finish it but we'll see) or integrated the DHT storage into the Satalite code see how it turns out. I will need to make the system so it will still work for Low-ID users and run on slow connections (I connect from a tethered mobile and apparently can only connect out). The good thing about Tor is it still works for me and in fact probably makes my connection faster and more reliable most of the time. I may make an entry socks server and exit node for it but its not really meant for that and I dont how to do SSL at the moment. I will also likely use UDP IP spoofing to hide the source of UDP packets, good for hiding Dos attacks but I thought it could also be used to hide communications aswell, but I assume this will likely not be an option in the future when firewalls get better at stopping them. Im aiming to have written as much of the code myself as possible, its all written in Delphi2 Im currently using Dave Burtons encryption and hashing code and Rudy an encryption I wrote. right now I cannot do and will need these for the final project (mainly cos I write my own code and I cant figure out how they work or I cant find any code for it on the net):- o Diffie-hellman. o RSA/El gamel public key encryption. (these will be pretty essential) o I can do Http but not SSL at the moment. o I dont have land line internet Once I got something that works I will also need someone on a fixed IP/domain to host a seednode for me so I can test it and people can boot into my network. Its just me working on this, not sure if I will finish this project but I will just keep working on it til I have something that works. Anyone interested email me. Comments/Feedback welcome. ~Shadowman. ~TheMindwareGroup themindwaregr...@gmail.com PGP: 0xf4b6586f -----BEGIN PGP SIGNATURE----- iQEcBAEBCgAGBQJS1yhbAAoJEKcLVST0tlhv90kIAKNWLX1lxwf3vhYrLRo0wpU+ duV4s0fA3zJoywLlAeyBPLC5KVD3cqpXlzaw2G+hItklNzYgrBtpPT11Zs86Mjnt jRExjdhg6w7a/DvOJRJBK4frBYIJTxcp1JtVyloja1f/EGZJ/DdGyuvMekCHWMhY 5pevBXbxPkY4VFNSMJ6cFIK3LgS5ZyFC9RlPXpJhkW2kPd2boR0QH97bri07muz9 CTtgLjaJeTCaJs3QYWsSNDG2kHx5OfdKpwO0Z/jvknNrCJCuA+xLXCHGtAnHho+9 gHxesRiH8B2vqeAVbiLx5lnorVJdYwqbuqHWwjUwb9bE1bk6OVMW3VGqisVjv6g= =R8f5 -----END PGP SIGNATURE----- -- tor-talk mailing list - tor-talk@lists.torproject.org To unsubscribe or change other settings go to https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk
