Settings

The table

MFTECmd and analyzeMFT both produce these columns. Keep the sub-second digits. An export truncated to whole seconds loses the strongest test on this page entirely, and the page will tell you when that has happened rather than quietly scoring everything lower. A master file table is a complete index of a machine; this one is read here and goes nowhere.

What the tests found

The four tests, and what each is worth

  • Dates flattened to exactly the secondNTFS keeps time in hundred-nanosecond ticks, so there are ten million of them in a second and a genuine timestamp essentially never lands on zero. The usual tooling sets times from a value accurate only to the second, which leaves the remainder at precisely nought. This is the strongest signal here, it does not depend on $FILE_NAME at all, and it is the one most tools never look at.
  • $STANDARD_INFORMATION earlier than $FILE_NAMEThe kernel writes $FILE_NAME when the file arrives, and nothing that happened to the file afterwards can have happened before it got there. $SI can be rewritten by any user-mode program with one documented call; $FILE_NAME cannot.
  • A file modified or read before it was createdImpossible, and the arithmetic does not care which tool made it so. Rare, but when it appears it is unambiguous.
  • A date far from its neighbours in the tableMFT records are handed out in order, so files created close together sit close together. This one is soft on purpose (records get reused when files are deleted) and is never a finding on its own.

The analysis


    

Questions people ask about TIMESTOMP

Why is the sub-second test stronger than comparing the two attribute sets?

Because it does not depend on anything the operator controls as easily. NTFS stores time in hundred-nanosecond ticks, so there are ten million possible sub-second values and a real timestamp lands on exactly zero about once in ten million. The common tooling sets a time from a value accurate only to the second, so the remainder comes out as precisely nought, and four dates on one file all landing there is not a coincidence in any meaningful sense. Comparing $SI to $FN, by contrast, is defeated by simply moving the file afterwards.

So can I use this to say a file was not tampered with?

No, and that is the most important sentence on the page. $FILE_NAME is written by the kernel, but the kernel writes it again every time a file is created, moved or renamed, so an operator who sets the dates and then copies the file into place leaves two sets that agree perfectly and sub-second digits that look entirely normal. Every test here catches the ordinary case and the careless case. Somebody who has read the same material as you will pass all four.

Where do I go when the answer actually matters?

The USN journal and $LogFile. Those record what was done to a file (created, renamed, data overwritten, attributes changed) with their own timestamps, rather than recording what the file says about itself. A file whose $SI claims 2021 but which appears in the journal being created last Tuesday is settled, and no amount of attribute forgery touches that. This page is the first ten minutes of that work, not the whole of it.

My export has no sub-second digits. What happens?

The page says so, in the figures and in the verdict, and scores nothing from that test rather than quietly marking everything as slightly less suspicious. It is a gap in the evidence, not a clean result, and it is worth going back for a better export: several tools truncate to whole seconds by default and that single setting throws away the strongest thing you had.

Why is the MFT record number test so weak?

Because record numbers are reused. When a file is deleted its record goes back on the pile and the next file created may take it, so a genuinely new file can sit among records from years ago perfectly innocently. It is still worth showing, because a backdated file dropped into a system directory often sits among records thousands of entries newer than the files it is pretending to be contemporary with, but on its own it means little, which is why it carries the least weight of the four.

What tools produce the input?

MFTECmd and analyzeMFT are the usual two, and both emit all eight timestamps as columns. Anything else that dumps $STANDARD_INFORMATION and $FILE_NAME side by side will do; you name which column is which rather than the page guessing, because every tool orders them differently and a guess that is wrong by one column would produce confident nonsense.

Doing this often? The Sovereign Console loads disk images and Windows artefacts into one case, and puts every timestamp on one timeline. On its own, US$2 a month; with everything else, US$5 a month or US$48 a year.