Two short, practical examples, which are gleaned from reality (though I am not at liberty to state of what organizations I speak):
One, educational institution. Due to compliance issues with student data being improperly smuggled out of the administration area (in reaction to an audit that found substantial lack of compliance with the educational record privacy laws), a comprehensive "there is no way that anything can leave without our knowing and tracing it" policy was put in place. USB memory sticks and cards were prohibited, and all outbound traffic was routed through a proxy. This proxy would behave normally in HTTP, and any HTTP CONNECT request was MITM'ed for recording purposes. Employees were notified that all of their machine and network activity were subject to logging, and were explicitly told that there was no expectation of privacy on the network. The MITMed connections were all given the same wildcarded security certificate, signed by an organizational CA which certificate was loaded into all browsers. (Technically, Nelson's proposal wouldn't cause a complete breakage, as long as there was sufficient horsepower and sufficient lack of controls available for the organizational CA to issue certificates at the request of the proxy server -- it could generate a key or use one of a pregenerated set of keys, send off the CSR with the name of the site that the client is sending the CONNECT request for, get an immediate return of the certificate signed by the organizational CA, load it, and then return that to the client. However, this goes very much against security best practices [a security-related operation performed automatically and without supervision by a CA with organization-wide trust at the request of an entity which is subject to attack from both internal and external threats].) Two, hospital. Due to HIPAA compliance issues, they must monitor all communications originating within their network, and as such they've put a proxy in place. One of their employees figured out how to to perform SSH over HTTP CONNECT; since this isn't appreciated by their MIS department, they broke any protocol negotiation other than SSL and TLS. Said employee didn't stop, though, and figured out a way to do an stunnel through which he ssh'es. This tunnel is unmonitored and completely opaque to the hospital MIS department. (He currently uses it to gain access to IRC and various talkers. While that employee is ethical and won't use it for anything which is expressly illegal... the fact that he can do what he's doing without detection means that others can do what /they/ want to do without detection. Including transmit "personally identifiable health information" without proper consent.) You'll run into the same issue in financial environments (fiduciary laws), public corporations (Sarbanes-Oxley and GLBA, aka the "Victoria's Secret" Law), health care providers (HIPAA), and all sorts of other places that need to ensure that communications are compliant with the law. MITM is unfortunately often necessary in these environments. -Kyle H On Dec 4, 2007 5:57 PM, Eddy Nigg (StartCom Ltd.) <[EMAIL PROTECTED]> wrote: > > Hi Kyle, > > I'm reading your mail now for the third time, but somehow I fail to > understand....Specially the thing about "Businesses/organizations need to be > able to enforce their policies". Could you give me a short, practical > example how the proposal from Nelson would break things and make it > impossible to enforce certain policies? I guess there might be minor > adjustments needed here and there, but not something which can't be overcome > in a sane amount of time...I really want to understand which use cases there > are which would support the current behavior. _______________________________________________ dev-tech-crypto mailing list dev-tech-crypto@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-tech-crypto
