Goblob is a lightweight and fast enumeration tool designed to aid in the discovery of sensitive information exposed publicy in Azure blobs, which can be useful for various research purposes such as vulnerability assessments, penetration testing, and reconnaissance.
Warning. Goblob will issue individual goroutines for each container name to check in each storage account, only limited by the maximum number of concurrent goroutines specified in the -goroutines
flag. This implementation can exhaust bandwidth pretty quickly in most cases with the default wordlist, or potentially cost you a lot of money if you're using the tool in a cloud environment. Make sure you understand what you are doing before running the tool.
go install github.com/Macmod/goblob@latest
To use goblob simply run the following command:
$ ./goblob <storageaccountname>
Where <storageaccountname>
is the target storage account to enumerate public Azure blob storage URLs on.
You can also specify a list of storage account names to check:
$ ./goblob -accounts accounts.txt
By default, the tool will use a list of common Azure Blob Storage container names to construct potential URLs. However, you can also specify a custom list of container names using the -containers
option. For example:
$ ./goblob -accounts accounts.txt -containers wordlists/goblob-folder-names.txt
The tool also supports outputting the results to a file using the -output
option:
$ ./goblob -accounts accounts.txt -containers wordlists/goblob-folder-names.txt -output results.txt
If you want to provide accounts to test via stdin
you can also omit -accounts
(or the account name) entirely:
$ cat accounts.txt | ./goblob
Wordlists
Goblob comes bundled with basic wordlists that can be used with the -containers
option:
- wordlists/goblob-folder-names.txt (default) - Adaptation from koaj's aws-s3-bucket-wordlist - a wordlist containing generic bucket names that are likely to be used as container names.
- wordlists/goblob-folder-names.small.txt - Subset of the default wordlist containing only words that have been found as container names in a real experiment with over 35k distinct storage accounts + words from the default wordlist that are part of the NLTK corpus.
- wordlists/goblob-folder-names.micro.txt - Subset of the small wordlist containing only words that have been found as container names in a real experiment with over 35k distinct storage accounts.
Optional Flags
Goblob provides several flags that can be tuned in order to improve the enumeration process:
-goroutines=N
- Maximum number of concurrent goroutines to allow (default:5000
).-blobs=true
- Report the URL of each blob instead of the URL of the containers (default:false
).-verbose=N
- Set verbosity level (default:1
, min:0
, max:3
).-maxpages=N
- Maximum of container pages to traverse looking for blobs (default:20
, set to-1
to disable limit or to0
to avoid listing blobs at all and just check if the container is public)-timeout=N
- Timeout for HTTP requests (seconds, default:90
)-maxidleconns=N
-MaxIdleConns
transport parameter for HTTP client (default:100
)-maxidleconnsperhost=N
-MaxIdleConnsPerHost
transport parameter for HTTP client (default:10
)-maxconnsperhost=N
-MaxConnsPerHost
transport parameter for HTTP client (default:0
)-skipssl=true
- Skip SSL verification (default:false
)-invertsearch=true
- Enumerate accounts for each container instead of containers for each account (default:false
)
For instance, if you just want to find publicly exposed containers using large lists of storage accounts and container names, you should use -maxpages=0
to prevent the goroutines from paginating the results. Then run it again on the set of results you found with -blobs=true
and -maxpages=-1
to actually get the URLs of the blobs.
If, on the other hand, you want to test a small list of very popular container names against a large set of storage accounts, you might want to try -invertsearch=true
with -maxpages=0
, in order to see the public accounts for each container name instead of the container names for each storage account.
You may also want to try changing -goroutines
, -timeout
and -maxidleconns
, -maxidleconnsperhost
and -maxconnsperhost
and -skipssl
in order to best use your bandwidth and find results faster.
Experiment with the flags to find what works best for you ;-)
Example
Contributions are welcome by opening an issue or by submitting a pull request.
- Check blob domain for NXDOMAIN before trying wordlist to save bandwidth (maybe)
- Improve default parameters for better performance
An interesting visualization of popular container names found in my experiments with the tool:
If you want to know more about my experiments and the subject in general, take a look at my article: