Wednesday, October 21, 2020

A Long Read With Something Bizarre

Review of Four FISA Applications and Other Aspects of the FBI’s Crossfire Hurricane Investigation

This Executive Summary excerpt isn't bizarre.

From the Executive Summary: 

Our review found that FBI personnel fell far short of the requirement in FBI policy that they ensure that all factual statements in a FISA application are "scrupulously accurate." We identified multiple instances in which factual assertions relied upon in the first FISA application were inaccurate, incomplete, or unsupported by appropriate documentation, based upon information the FBI had in its possession at the time the application was filed. We found that the problems we identified were primarily caused by the Crossfire Hurricane team failing to share all relevant information with OI and, consequently, the information was not considered by the Department decision makers who ultimately decided to support the applications.

As more fully described in Chapter Five, based upon the information known to the FBI in October 2016, the first application contained the following seven significant inaccuracies and omissions:  

1.Omitted information the FBI had obtained from another U.S. government agency detailing its prior relationship with Page, including that Page had been approved as an "operational contact" for the other agency from 2008 to 2013, and that Page had provided information to the other agency concerning his prior contacts with certain Russian intelligence officers, one of which overlapped with facts asserted in the FISA application; 

2.Included a source characterization statement asserting that Steele's prior reporting had been"corroborated and used in criminal proceedings," which overstated the significance of Steele's past reporting and was not approved by Steele's handling agent, as required by the Woods Procedures; 

3. Omitted information relevant to the reliability of Person 1, a key Steele sub-source (who was attributed with providing the information in Report 95 and some of the information in Reports 80 and 102 relied upon in the application), namely tha

 ( 1) Steele himself told members of the Crossfire Hurricane team that Person 1 was a "boaster" and an "egoist" and "may engage in some embellishment" and 

 (2) the FBI had opened a counterintelligence investigation on Person 1 a few days before the FISA application was filed;  

4. Asserted that the FBI had assessed that Steele did not directly provide to the press information in the September 23 Yahoo News article based on the premise that Steele had told the FBI that he only shared his election-related research with the FBI and Fusion GPS, his client; this premise was incorrect and contradicted by documentation in the Woods File-Steele had told the FBI that he also gave his information to the State Department; 

 5. Omitted Papadopoulos's consensually monitored statements to an FBI CHS in September 2016 denying that anyone associated wit h the Trump campaign was collaborating with Russia or with outside groups like Wikileaks in the release of emails;  

6. Omitted Page's consensually monitored statements to an FBI CHS in August 2016 that Page had "literally never met" or "said one word to" Paul Manafort and that Manafort had not responded to any of Page's emails; if true, those statements were in tension with claims in Report 95 that Page was participating in a conspiracy with Russia by acting as an intermediary for Manafort on behalf of the Trump campaign; and  

7. Included Page's consensually monitored statements to an FBI CHS in October 2016 that the FBI believed supported its theory that Page was an agent of Russia but omitted other statements Page made that were inconsistent with its theory, including denying having met with Sechin and Divyekin, or even knowing who Divyekin was; if true, those statements contradicted the claims in Report 94 that Page had met secretly with Sechin and Divyekin about future cooperation with Russia and shared derogatory information about candidate Clinton. None of these inaccuracies and omissions were brought to the attention of OI before the last FISA application was filed in June 2017. Consequently, these failures were repeated in all three renewal applications. Further, as we discuss later, we identified 10 additional significant errors in the renewal applications.

 I have not read all 434 pages.  But I bring this to your attention because a friend noticed something odd and disturbing.  He found when examining a previous issue that searching for "Comey" failed to find many occurrences of Comey's name.  In this revision, I found 12 places where searching for "Corn" matched Comey's name but more than 50 places where "Comey" matches.  It appears that the "Corney" matches were the result of kerning the r and n into something that looks like "Comey", but cannot be found by searching.  Kerning is a feature of Microsoft Word  (look under Format Font Advanced) but this would have to be done individually for each such occurrence.  It makes it hard to find all uses of Comey's name.  My informant says the previous issue had even more such cases.  Can anyone explain this weirdness?


  1. Would the r + n kerning job be possible with a "search and replace" ?

  2. I've seen this sort of thing in many, many PDF files. It is caused by the original source being a scanned image which is run then through OCR without correcting any of the inevitable errors.

  3. I've seen things like that occur with optical character recognition from scanned documents, but why there would be scanning involved these days is anyone's guess. My low opinion of government's idea of efficiency wouldn't rule it out though.

  4. Hal & RevGreg: For a 1970s document, they might have scanned it. All current Word documents would be exported as a PDF.

    J Melcher: That would be readily done with a keyboard macro.

  5. The number of documents which should have been printed directly to PDF but instead are printed then scanned is truly astonishing, and it's not just government that does it. At least with mostly-text documents, the OCR keeps the file size somewhat reasonable. Construction plan sets are horrible - sheets that are a quarter MB printed as PDF are 3 MB when scanned. And even small projects will have hundreds of sheets.

    There's a term for when the spacing between letters isn't right and makes it difficult to read: keming.

  6. Clayton, your humble informant here.... In the Dec. 11th version (first version was Dec. 9th, which I did not see) every single instance of Comey's name was spelled "Corney," with kerned r + n. If memory serves, they were spread across 65 pages of the pdf. Before I wrote you about this the other day I went through my copy of the Dec. 11th version just to verify that that was the case. The next released version had all but two instances corrected to "Comey" with an m.

    And I agree with you that it's highly unlikely that any OCR process was involved in the preparation of this document, not here in 2019.

  7. Only pages 1, 2, and 478 are text. Pages 3 through 477 are scanned images. If you zoom in on any one of those pages, you can clearly see the artifacts from the scanning process.

    The PDF was created with Adobe Paper Capture Plug-in.

    "You can use Adobe Acrobat with a scanner to create Adobe PDF files from paper documents. The resulting file is a PDF Image Only file—that is, a bitmap picture of the pages that can be viewed in Acrobat but not searched.

    "If you want to be able to search, correct, and copy the text from an Adobe PDF Image Only file, you can “capture” the pages in three file formats. Adobe PDF Formatted Text and Graphics, PDF Searchable Image (Exact) and Searchable Image (Compact) all apply optical character recognition (OCR) and font and page recognition to the text images and convert them to normal text. The Searchable Image file types have a bitmap image of the pages in the foreground, and the captured text on an invisible layer beneath it"