Sat May 17 10:29:54 PDT 2014

Redundancy: Data history redundancy: How many copies of data history should be retained, where, and for how long?


Options:

Option 1: Use a single data history repository
Option 2: Use at least two data history repositories
Option A: Co-located with the plant.
Option B: At multiple locations.
Option C: At multiple locations adequately distant for the common modes of import.
Option D: Where otherwise convenient.
Option U: Until not required for normal business uses.
Option V: Until regulatory mandates are fulfilled.
Option W: Until liability is mitigated.
Option X: Until forensic images made and secured.
Option Y: Until after action reports and related matters are settled.
Option Z: Until identified operational uses end.

Decisions:

Use all rows that apply.

Threat Consequence Copies / Where / How Long
High OR Insider Med+ How many: At least two.
Where: At multiple locations.
How long: Until not required for normal business uses.
-- Regulatory Violations How many: At least two.
Where: At multiple locations.
How long: Until regulatory mandates are fulfilled.
-- Excess liability, recalls How many: At least two.
Where: At multiple locations.
How long: Until liability is mitigated.
Intentional spoliation or alteration Med+ How many: At least two.
Where: At multiple locations.
How long: Until forensic images made and secured.
Common mode failures Med+ How many: At least two.
Where: At multiple locations adequately distant for the common modes of import.
How long: Until after action reports and related matters are settled.
-- -- How many: One.
Where: Co-located in plant or where otherwise convenient.
How long: Retained until identified operational uses end.
How many copies of data history should be retained, where, and for how long

Basis:

The alternatives come down to Use [one | two | many] data history repositories located [with | separated from] the ICS and each other, and retained until [what condition].


How many:

Use a single data history repository well protected from all identified threats.
For most plants, a single data history repository is adequate because the only use of the history is to check on process failures and improve performance.

Use at least two data history repositories.
If forensics is going to be performed, forensic soundness mandates that a copy be made and securely stored. This can be done anywhere, as long as the storage is adequate to the forensic need. Normally, it is necessary to assure that such evidence is not altered or spoliated, and to consider that the legal purpose may drive those motivated for one purpose or another to alter the data history. Thus separation from other copies and all interested parties is required, as is a chain of custody associated with the data history to be used in court.


At multiple locations.
If there is only one copy, multiple locations will not be required or possible. Therefore, they are only relevant when there are multiple copies.

If the data history is used to detect and/or react and adapt to intentional abuse, then there is a need to prevent the alteration of the data history by intruders. This would seem to mandate an append-only storage process and a read-only access process. Redundant copies kept elsewhere are thus important in cases where insiders may be involved or threats are high.

Records of production details are often required for systems that ultimately effect human safety, such as safety-related automotive parts, systems used to control food and drug production, and so forth. In these cases, redundancy may be necessary to assure use in recalls and after incidents, and redundancy may be used to increase the assurance associated with these cases. Over time, separate and different mechanisms will be required to afford high surety.

In cases where plant events or other common mode failures can destroy or make inaccessible a data history repository, redundant copies may be required at remote locations both to perform after-action reviews and to support legal and regulatory mandates.


Kept until when?
Typically, a data retention and disposition policy mandates the retention requirements on data history. To the extent that no such policy review has been undertaken, the following minimum retention time guidelines are a starting point:


In all cases: Data history repositories should be tested and verified periodically. This is typically done at least once per year as part of standard business processes. History and redundant copies of history should be verified as they are made to detect process errors as soon as feasible.

Since historical data is rarely time critical, there is no particular need to have live backup copies, and standard backup mechanisms may be adequate. If data history repositories are properly secured to allow append only access from ICS systems and read-only access for auditors, forensic examiners, and engineers seeking to debug and improve processes, no substantial value comes from additional redundancy.

Copyright(c) Fred Cohen, 1988-2012 - All Rights Reserved