Hi ! First of all, I am sorry for typos/bad spelling sometimes. >I think you're heading down the wrong path, my friend. You're doing way >more work than you need. Before asking how to implement such a detailed >process (which seems more like a patchwork of random phrases from Applied >Cryptography), you need to define exactly what you're trying to protect, >who you're trying to protect it from, and who your users are. Furthermore, >your description sounds like a reinvention of HTTP digest authentication, >among other things. I tried. I can't explain better..... Basically, in one sentence, I can say: I need to protect HTML thin client of busines system. It needs to be accessed from untrustworthy terminals - such as hotels, customers, etc. Because of this, I need something, what doesn't store ANYTHING on terminal or require INSTALATION. We have to assume that keyboard is logged, connection sniffed. The result of authentication (I hope I spelled well :o) is enabling access to bysines system (separate domain). That's all. As I wrote, I prefer some 'allready done way' - but I didn't found any, which meets this simple needs. I looked around RSA etc. but I didn't found (anyway: we allready have SecurID related products - but for different pusposes). What will fit our needs ? >Keep in mind that proprietary security solutions are very rarely found to >be secure, once tested by experts. Using "some weird CPU's" is not a valid >reason for avoiding established algorithms and protocols. Open, >established systems for crypto and authentication already exist, and are >used every day by people with very strong security requirements. Standards >like AES have been pounded, beaten, smashed, and run over by the top crypto >experts in the world. I think you're misled in thinking you have to start >from scratch. I didn't find anything, what fit our needs. If there is something, it requires INSTALATION or it runs only on WINDOWS. And I use "wierd CPU" only because I didn't find better solution than generation floating access code at different machine, to supress storing datas on host computer/instalations -> small separate machine (with some CPU generating access codes). And anyway, similar principle uses many banks - and they prefer it BEFORE asymetric cryptography [and they are security experts]. They give you small 'calculator' -> and it works similar way as I described (a bit different due different purposes). It allows you to login into HTTPS banking, wire anything, standing ordering, create new accounts, ... Basically everything - you don't need to go into bank anymore. Is this way (in general) secure then ? And about 'security exprests': yes, we should. We allready did it for few things (network based communications - we are HW developers). We never had problem yet. I belive in simple things. And it is verry hard to find 'real' security experts anyway - majority of "experts" only plays on it :( I personally prefer to make this small thingy open source (if we will do - I say many times, that I am going to do that only because I don't see different solution), maybe someone will need it too, look at it, etc. Real security experts doesn't work at security companies IMHO. >By the way, spell check is your friend. "Authentication" and >"cryptography" don't contain F's. I am sorry. I will try to use proper spelling. R.
This archive was generated by hypermail 2b30 : Tue Apr 23 2002 - 13:22:01 PDT