Bhagat - I disagree with a number of your comments - my text inline below. Todd ----- Original Message ----- From: "Devdas Bhagat" <devdasat_private> To: "Log Analysis Mailing List" <loganalysisat_private> Sent: Tuesday, December 04, 2001 10:49 AM Subject: Re: [logs] Due Diligence for Admission in Court > On 04/12/01 09:40 -0800, todd glassey wrote: > > If you plan on submitting something to a court, you likely may need to have > > the following setup: > > > > 0) Understanding what it is you are trying to do > Important point, missed this one. > <snip> > > > > 1) Some proof of the time data that was stamped on the document. > > > > This is a real issue since NTP across the open Internet is not reliable > Ok, but you can setup your own local NTP server. If your local time > stamps are consistent across multiple local systems, you have much less > to worry about. > For multiple locations, one local NTP server and local log server, > plus a central log server and central NTP server. > Reasonable attempts to maintain timestamp synchronization. the question is how to get time to that NTP Server in the form of an Initialization Event and developing a Timebase management process. > > <snip> > > > 2) Some level of operating integrity which means a regularly run IDS > > process to verify the consistency of the system, like AIDE, SHADOW, or any > > of the commercial ones (Tripwire, ISS, etc). > > Yes, logging and log checking would be a part of this process. > > > The bottom line is that you need to prove that your environment was behaving > > properly and to do that you will need these Filesystem IDS tools. The IDS > > Actually, you need to prove that it had integrity before the compromise. > Behaving properly, or not is a different term and will depend on what > you want it to do. That's the point. The audit must take into account what the system was supposed to be doing and what not. IDS processes need to start long before any breakin or compromise can possibly occure to set a baseline for the oeprations of the specific system under logging protection. > > <snip> > > 4) Some kind of Log Management regimen wherein the logs are regularly > > timestamped, rotated, and made tamper-proof > > Again, see point two. You need to show integrity. Thats the only thing > you need to prove. No its not. You need to get more into timestamping to see where the holes in this statement are. > > <snip> > > 5) Some kind of active audit model that demonstrates that each one of > > these constraints or subsystem requirements are met. > > > > This is likely between you and your auditors who ever they are and then your > > data and process will likely be admissible just about anywhere. > > As long as a consistent audit model is maintained, and well documented. > If the model is not documented, it will not be admissible. I disagree with this as a blanket statement with no founding in any court of law that I am familiar with. If the model can be proven by audit to be consistent then the documentation is a nicety. > > <snip> > > > For example, you can log the SMTP sender, but since that is in the hands > > > of the client, it cannot be verified, or trusted. > > > > Only true if Sendmail/mailx is used. Many of the other mailer agents have > > addressed this and allow for extended handshaking. > > What the client supplies cannot be trusted. Only locally > originated/locally verifiable data can be trusted. This simply is untrue. If the system does not supply SMTP on port 25 then this whole question is moot. It also is an assumption of the clients operating criteria, and that is unaccepetable in any audit model. > > > > Even if you restrict only authorized clients writing to the datastream, > > > you have no knowledge that the data itself is valid. > > > > You are not worried as to the integrity of the data only the constency of > > the data once it entered your system. > > You have to worry about both. If you cannot trust the integrity of the > data itself, you have problems with your tools, since they have to trust > this data to act on it. Simply not true - The application that is using the data needs to know that the data is OK, the logging system doesn't care. Remember that the logging system is part of the proofing model and not the decision support systems that makes up a part of the application. > GIGO. > If your data is not consistent, it should immediately ring alarm bells. yes but this is about interprtation and an operating model, not a capability. Note that you are mixing MUST's/SHOULD's which have to do with process and other constructs that are user selectable with operating processes, and this is a mistake I think. > If the data is consistent, then its integrity has to be validated in some > form, as far as possible. Only if one cares what the data is. Many processes, especially logging ones will not. The applications that log to them might, but that is a different layer of the puzzle. remember there are two purposes of information in a system, one is in support of some decision support process and the other is evidentiary in nature. > > > > > > > > - lack of time synchronization > > > NTP, ideally following the new RFC. > > > > No No No - NTP is not reliable over the open Internet - period. > See what I said above. Expect that errors will occur. Make reasonable > attempts to keep the errors to a minimum. > > The operative word here is reasonable. I will not expect it reasonable > that a typical home user needs to have the expertise needed to configure a > firewall, but I would expect a much higher degree of competence from a > professional administrator. > <snip> OK but can they use a dialout client to set their time from NIST ACTS? > > Devdas Bhagat > > --------------------------------------------------------------------- > To unsubscribe, e-mail: loganalysis-unsubscribeat_private > For additional commands, e-mail: loganalysis-helpat_private > > --------------------------------------------------------------------- To unsubscribe, e-mail: loganalysis-unsubscribeat_private For additional commands, e-mail: loganalysis-helpat_private
This archive was generated by hypermail 2b30 : Tue Dec 04 2001 - 14:02:10 PST