110 likes | 258 Views
“A general strategy for differential forensic analysis”. Presented by: Garrett Leach. http://www.dfrws.org/2012/proceedings/DFRWS2012-6.pdf. Quick Definitions. Image – “A byte stream from any data-carrying device” Feature – Information implied or extracted from image.
E N D
“A general strategy for differential forensic analysis” Presented by: Garrett Leach http://www.dfrws.org/2012/proceedings/DFRWS2012-6.pdf
Quick Definitions • Image – “A byte stream from any data-carrying device” • Feature – Information implied or extracted from image
Prior Work (And related) • We are already familiar with diff • File/Data sync programs • Rsync • Git • Svn • Previous timeline reconstruction papers
Use cases • Malware discovery • Discover alterations to system files and registry • Insider Threat Detection • Discover irregularities in time • Generalizable to other areas (Pattern of life) • Pre-emptive routine imaging* • Discard files that will not be important • Windows update files
The generalized approach • Collect Feature Metadata • Location(s) • Name(s) • Timestamps • Create list of changes • Locate inconsistencies in time • Beware common pitfalls • Report
Report Generation • Intent: Suppress unwanted features • Common techniques: • Present statistics (instead of enumerating features) • Organize features into hierarchies that can be expanded (folder, file/name, metadata) • Organize features into timeline
Tools created • idifference(.py) • Reads DFXML (for each image) • Outputs filepaths in 1st image and not second, inodes added to 2nd, and filepaths added to 2nd • rdifference.py • Reads two windows registry files • Outputs: new and deleted cells and values (both content and type) and keys with changed mtimes • Caveat: Some cells have non-unique paths
Tools created • bulk_diff.py • Compares data (histograms) between runs of bulk_extractor (from another paper) • corpus_sync.py • Sync program that recognizes renames • Not deletion and creatioon • flowdiff (in-progress as of publication) • Processes pcap files to create DFXML