Document marking using alternative terms for document source detection
Unique copies of an original document can be generated and provided to individual recipients. The unique copies can be used to identify the source of a document leak. The unique copies are generated by replacing terms within the original document with alternative terms. The alternative terms are determined using a first machine learning model that receives a term from the document and outputs the alternative terms. The output alternative terms are provided to a second machine learning model that indicates a tone for each alternative term. The tone of the alternative terms is compared to the tone of the term from the original document, and one or more of the alternative terms are selected based on the tone of the alternative terms relative to the tone of the document term. The alternative terms used to generate the unique copies have a same or similar tone as the document term.
1 . A system for document source detection using unique copies, the system comprising:
at least one processor; and
one or more computer storage media storing computer readable instructions thereon that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
accessing a document comprising terms;
determining a set of alternative terms for a term in the document, the set of alternative terms comprising a first alternative term and a second alternative term;
generating at least two different unique copies of the document for distribution to different recipients, the unique copies comprising a first unique copy generated by replacing the term with the first alternative term and a second unique copy generated by replacing the term with the second alternative term; and
indexing an association between the unique copies and the recipients of the unique copies, such that the index is usable to identify a recipient for an identified unique copy.
2 . The system of claim 1 , further comprising identifying a frozen term within the document using a term index, wherein the frozen term is kept constant when generating the at least two different unique copies, the frozen term being included in the term index based on a previous selection made within a different document for which unique copies were generated.
3 . The system of claim 1 , further comprising:
receiving an artifact; and
determining a source of the artifact from the different recipients based on comparing terms of the artifact with terms from the unique copies.
4 . The system of claim 1 , further comprising:
presenting the set of alternative terms at a graphical user interface; and
when generating the at least two different unique copies, replacing the term with the first alternative term and the second alternative term based on a selection from the presented set of alternative terms.
5 . The system of claim 1 , further comprising:
receiving a selection of terms within the document; and
replacing the term with the first alternative term and the second alternative term based on the term being excluded from the selection of terms.
6 . The system of claim 1 , wherein the set of alternative terms is further determined based on an author of the document using a model trained on a set of documents by the same author.
7 . The system of claim 1 , further comprising selecting the first alternative term and the second alternative term based on a number of words within the term.
8 . One or more computer storage media storing computer readable instructions thereon that, when executed by a processor, cause the processor to perform a method of document source detection using unique copies, the method comprising:
accessing a document comprising terms;
determining a first set of alternative terms for a first term in the document and a second set of alternative terms for a second term in the document;
generating at least two different unique copies of the document for distribution to different recipients, the unique copies comprising a first unique copy generated by replacing the first term with a first alternative term selected from the first set of alternative terms and a second unique copy generated by replacing the second term with a second alternative term selected from the second set of alternative terms; and
indexing an association between the unique copies and the recipients of the unique copies, such that the index is usable to identify a recipient for an identified unique copy.
9 . The media of claim 8 , further comprising:
receiving a selection of terms within the document; and
replacing the first term and the second term in the document based on each being excluded from the selection of terms.
10 . The media of claim 8 , further comprising identifying a frozen term within the document using a term index, wherein the frozen term is kept constant when generating the at least two different unique copies.
11 . The media of claim 8 , further comprising:
receiving an artifact; and
determining a source of the artifact from the different recipients based on comparing terms of the artifact with terms from the unique copies.
12 . The media of claim 8 , wherein the first set of alternative terms and the second set of alternative terms is determined based on an author of the document.
13 . The media of claim 8 , further comprising selecting the first alternative term based on a number of words within the first term and selecting the second alternative term based on a number of words within the second term.
14 . The media of claim 8 , further comprising:
presenting the first set of alternative terms and the second set of alternative terms at a graphical user interface; and
when generating the at least two different unique copies, replacing the first term with the first alternative term and the second term with the second alternative term based on selections from the presented first set of alternative terms and the presented second set of alternative terms.
15 . A computer-implemented method of document source detection using unique copies, the method comprising:
accessing a document comprising terms, including a first term and a second term;
distributing a first unique copy of the document to a first recipient, the first unique copy having a first alternative term replacing the first term, the first alternative term being selected from a first set of alternative terms determined using the first term; and
distributing a second unique copy, that is distinct from the first unique copy, of the document to a second recipient, the second unique copy having a second alternative term replacing the second term, the second alternative term being selected from a second set of alternative terms determined using the second term.
16 . The method of claim 15 , further comprising:
receiving a selection of terms within the document; and
replacing the first term with the first alternative term and the second term with the second alternative term based on the first term and the second term being excluded from the selection of terms.
17 . The method of claim 15 , further comprising:
receiving an artifact; and
determining a source of the artifact from the different recipients based on comparing terms of the artifact with terms from the unique copies.
18 . The method of claim 15 , further comprising:
presenting the first set of alternative terms and the second set of alternative terms at a graphical user interface; and
replacing the first term with the first alternative term and the second term with the second alternative term based on selections from the presented first set of alternative terms and the presented second set of alternative terms.
19 . The method of claim 15 , further comprising:
identifying a frozen term within the document using a term index; and
generating the first unique copy and the second unique copy such that the frozen term is constant between the document, the first unique copy, and the second unique copy.
20 . The method of claim 15 , further comprising
generating the first unique copy by replacing the first term with the first alternative term based on a tone of the first term and the first alternative term; and
generating the second unique copy by replacing the second term with the second alternative term based on a tone of the second term and the second alternative term.