This Assemblyline service extracts metadata and objects from PDF files using Didier Stevens PDFId (Version 2.7) and PDFParser (Version 7.4) tools.
NOTE: This service does not require you to buy any licence and is preinstalled and working after a default installation
- ADDITIONAL_KEYS: List of keywords searched for by PDFid
- HEURISTICS: Choose the heuristics plugin to be run during the service execution. Here's the list of plugin to choose from:
- pdf_id/pdfid/plugin_embeddedfile
- pdf_id/pdfid/plugin_nameobfuscation
- pdf_id/pdfid/plugin_suspicious_properties
- pdf_id/pdfid/plugin_triage
- MAX_PDF_SIZE: Maximum size PDF to be processed by PDFid. This value will be ignore during deep scan
The PDFId service will report the following information for each file when present:
- PDF Header String
- Number of:
- objects
- streams
- endstreams
- xref
- trailer
- startxref
- '/Page'
- '/Encrypt'
- '/Objstm'
- '/JS'
- '/Javascript'
- '/AA'
- '/OpenAction'
- '/AcroForm'
- '/JBIG2Decode'
- '/RichMedia'
- '/Launch'
- '/Colours'
- '%%EOF'
- Bytes after %%EOF
- Total entropy
- Entropy inside streams
- Entropy outside streams
- Mod Date (AL tag: file.pdf.date.modified)
- Creation Date (AL tag: file.date.creation)
- Last Modification Date (AL tag: file.date.last_modified)
- Source Modified Date (AL tag: file.pdf.date.source_modified)
Note: PDFParser will only run on a sample if in deep scan mode, or if PDFId plugins (see below) detected suspicious elements are present in the sample.
-
Reports number of:
- /Comment
- /XREF
- /Trailer
- /StartXref
- /Indirect object
- /Catalog
- /ExtGState
- /Font
- /FontDescriptor
- /Pages
-
Extracts PDF Elements:
- Comments
- Trailer
- StartXref
-
Extracts Suspicious Elements:
- Entire Objects (as extracted file) when flagged by PDFId plugins (JBIG2Decode objects will only be extracted in deep scan mode).
- Specific Object content (in AL result) and will run FrankenStrings Patterns against content to search for IOCs (determined by PDFId plugins).
-
ObjStms
- Service will attempt to reprocess object streams in samples as PDF files to re-run against PDFId and PDFParser analyzers. If in deep scan mode, a maximum of 100 objstms will be reprocessed, otherwise a maximum of two will be reprocessed.
PDFId plugins are python scripts used by PDFId service to score suspicious properties based on PDFId results. Plugins can be added to service by users (see configuration above). The following format is required for plugin scripts to work with this AL service:
class cPDFiD[NAME](cPluginParent):
onlyValidPDF = True
name = '[NAME OF PLUGIN]'
def __init__(self, oPDFiD):
self.oPDFiD = oPDFiD
# Whether or not hits is used, it must be returned by Score
self.hits = []
def Score(self):
score = 0
[conditions that might adjust score/self.hits]
return score, self.hits
def Instructions(self, score, hits):
if score == 1000:
# These messages will show in AL result,
along with score
return 'Some message'
if score == 500:
return 'Some other message'
if score == 0:
return
See source code under "pdfid" folder for examples of plugins already used by this service.