Fable
Inserting notes and altering the manuscript
Old Shen, the director of the archive, encountered two strange things. First: a researcher came to the archive every day to look up materials, and one day he found two forged excerpts secretly slipped onto his reading desk. He wrote his research report based on the fake excerpts, and all conclusions were skewed. The next day he switched desks, the fake excerpts disappeared, and subsequent reports returned to normal.
The second matter was more serious. Warehouse management discovered that the original files in the archive room had been tampered with—someone had replaced key passages in several critical documents with forged content and re-filed the modified versions. From then on, every researcher who came to consult that file received false material.
Old Shen set out to investigate. The note-slipping incident was quickly solved: every day before the library opened, a cleaner came in and out of the reading room, and someone slipped in while she was not paying attention to tamper with things. The solution was to add access control to the reading area and have researchers clear their desks when they left. The cost was small, and the fix worked immediately.
The draft-tampering incident dragged on for two months. Old Shen had to compare the original versions and current versions scroll by scroll, and after finding the tampered files, he still had to retrieve the original scans from the backup repository and restore them one by one. The most troublesome part was that during those two months, research reports, policy recommendations, and investment decisions produced based on the false files had already been distributed to more than a dozen departments.
Old Shen distinguished between two kinds of events. Slipping in a note is "session-level interference": the contaminant enters the reading table, affects that consultation, and clearing the table makes it clean. Changing the draft is "knowledge-base-level destruction": the contamination enters permanent storage, all subsequent consultations are contaminated, and the cleanup cost grows exponentially with the spread.
The archive later set two rules. First, all original files were compared with the backup repository once a month, with immediate alerts for any discrepancy. Second, researchers had to attach the file numbers and version numbers they cited when submitting reports, so problems could be traced back.
At an industry exchange meeting, Old Shen said: to deal with slipped-in notes, managing the access zone is enough. To deal with altered drafts, you need backups, comparison and traceability. The defense investment for these two problems differs by an order of magnitude, but many people treat them as the same thing.
After the meeting, someone asked Old Shen which attack was the most dangerous. Old Shen said: the most dangerous is when you think you are defending against altered drafts, but in fact you have only added protection against slipped-in notes. After half a year of calm, you suddenly discover thirty fake case files hidden in the warehouse—that is what it means to be too late.