Didier Stevens

This update to jpegdump.py, my tool to analyze JPEG images, brings 2 small changes:

Data between segments can be selected with suffix d. Like this: -s 10d

This means: select the data between segments 9 and 10.

And when option -E is used to add hash values, repeating hashes are marked with parentheses.

jpegdump_V0_0_10.zip (http)
MD5: 5B33C0ECB94E3284CA64E98B5A0947C3
SHA256: D8C657DB7564160725C95677BE200EB3A902BDC74CF335EFA8499596495633F0

This is a bug fix version and also adds updated statistics.

1768_v0_0_16.zip (http)
MD5: E72E66BE5A66DC2C6E1806DE82DF9B39
SHA256: 008E15C617EE94D849A3325643497D216E559609602E97CF2EE41968CCA5D096

Some new features that help with analyzing memory dumps.

Here is the analysis of a VMware vmem file:

There’s a new sanity check, determining if an extracted configuration is OK or not OK (NOK).

A config passes the sanity check if it contains a valid payload type and a valid public key.

Configurations that don’t pass the sanity check, are most likely false positives: they have a valid header, but no valid fields. They can show up in memory dumps of Windows machines.

Option -S can be used to hide configurations that don’t pass the sanity check:

Now we are just left with detections of the sleep mask routine. What’s new in this version, is that the position where the signature was found is listed.

Finding both 32-bit and 64-bit routines is unusual.

Option -V can be used to dump 256 bytes before and after the signature, to help us get an idea what we are dealing with.

And what we actually found here, is the memory of the anti-virus program containing signatures, like signatures for Cobalt Strike sleep mask deobfuscation routines.

1768_v0_0_15.zip (http)
MD5: 15EBA21D59D78ED9A674DC2B88687555
SHA256: 73987F1B8577A5C31B2D7BDC197A465F8700B3F3C7838A31802BD77FFC872C42

This new version of re-search.py adds a regex for UNCs to the library and has a Python 3 fix.

re-search_V0_0_21.zip (http)
MD5: 294DD5D4027F0AFD0A2DE6432FE4552D
SHA256: B818CE4F7E217B381128550A3A36B40B6D07CC687CE4CF5AFF3C70EC0D3EEAD2

This update brings an update to plugin plugin_vba_dco.py.

This is a plugin that scans VBA source code for keywords (Declare, CreateObject, GetObject, CallByName and Shell), extracts all lines with these keywords, followed by all lines with identifiers associated with these keywords.

For example, if the result of a CreateObject call is stored in variable oXML, then all lines with this oXML identifier are selected.

I updated this plugin with two options -g (–generalize) and -a (–all).

Option -g generalize will replace all identifiers (like variable & functions names) with a general name: Identifier#### where #### is a numeric counter.

I added this option to analyze a sample where almost all identifiers where completely unreadable, as they consisted solely out of characters that are between byte values 128 and 255 (e.g., non-ASCII).

Here is the output for that sample, without using any plugin option:

You can see the CreateObject functions, but appart from the WshShell identifier, the other identifiers don’t have letters and are hard to trace in the code.

This changes when you use option -g:

All identifiers have been generalized to names like Identifier0001, Identifier0002, …

To view all generalized code (and not only the lines with keywords), use option -a:

Remark that this plugin is not a VBA parser: it uses some simple scans and regexes to find identifiers. For example, it handles line comments like any other lines.

oledump_V0_0_69.zip (http)
MD5: 9FDE05EB0B475C5BB76A92A926DBE8CD
SHA256: 16761C633DEC83CB691AE7223BB5AE82E5EC668F5D161499800638BC45420285

This new version adds JSON input support, allowing,for example, to detect encoded payloads inside the registry:

More info in an upcoming blog post.

base64dump_V0_0_23.zip (http)
MD5: 00D1E2344A6D09D3A2F18FC257F77090
SHA256: E4CA046198E801DFF309D6A8B346D5084FB4B4DFBFD339C5BCB3EF570CD08A79

This new version of format-bytes.py adds a feature to search for a range of integers:

#iv5#6080 means: look for an integer (i) equal to 6080 with a variation of 5 (v5), e.g., look for integers between 6075 and 6085.

format-bytes_V0_0_14.zip (http)
MD5: 600969FAC1F397036673574EA0BE0EE1
SHA256: D0EB0709985A4A5FEC1DA4B420CA440FF5268229CFFA1B3CC1EE5FAE92101957

This new version contains a Python 3 fix.

cut-bytes_V0_0_15.zip (http)
MD5: 1906873950C1DC55665072C7F3529D7F
SHA256: 2B9847E49C08021C61B8FA09C9DD400FC41E817F65E1C2BAC64ABBD87D49E238

This new version of base64dump.py adds some extra info for the encoded strings.

In -e all mode, a new column Chars tells you how many unique characters are used for that encoded string:

For example, the last line is recognized as a syntactically valid variant of BASE85 (b85), but it uses only 63 unique characters (85 unique characters is the maximum). So this is probably not b85, or else the encoded data has low entropy.

And there is also new info when you select a string for info:

base64dump_V0_0_22.zip (http)
MD5: B38E4F454FAE219D771742B44D60A428
SHA256: 32695EEDDADAE9B1AFA1CAA70A69E2A0434E2264CEF836DE172BC5254C8E6281

This new version adds option -l to provide a short list via an option, in stead of using a file. And there’s a Python 3 bug fix.

python-per-line_V0_0_8.zip (http)
MD5: C7A61FE8FF701BC3A49CF7C093FB290D
SHA256: 63AEBD847D26A9B25F401D8734FBED646E7BB3F9DF2238EF49ACEAB2E1EF5AFA