[Google] Improper Data Deletion for Google AI Studio Prompts

 Google AI Studio database saves deleted prompts made by its users (not the LLM), resulting in privacy issues as follows -

* Google is saving data which has been deleted by its users against their choice.

* Users are in a false knowledge that these prompts are deleted. Users don't have control over their own data thrown in prompts.

* A public prompt (shared prompt) will always be available to public view (or selected view according to users sharing settings) even after deletion.

Timeline -

Reported - 28.05.2024

Closed - 28.05.2024 as "Not a security issue".

Popular posts from this blog

[Google] Access to BGP server + DOM XSS

 A hacker could have had the access to a BGP REST API server without any authorisation which was leading to leak of internal network info, devices info, network configuration and modification. Apart from that, there was a DOM XSS vulnerability in the target IP asset as well. POC - An IP address had a BGP server running on it. The access to it was an API service " Sonic Network Management APIs". These API calls could have been made without authorisation(access_token) and they were all listed with specifications on - " https://IP/ui/model.html". Endpoints such as - /ui/model.html?urls.primaryName=sonic-interface.yaml#/sonic-interface/get_sonic_interface_sonic_interface /ui/model.html?urls.primaryName=sonic-port.yaml#/sonic-port/get_sonic_port_sonic_port /ui/model.html?urls.primaryName=openconfig-lldp.yaml#/openconfig-lldp/get_openconfig_lldp_lldp_interfaces /ui/model.html?urls.primaryName=openconfig-platform.yaml#/openconfig-platform/get_openconfig_platform_components...
Read more

[Google] YouTube "restconf" Swagger-UI XSS

  During my recon, I came across a few Google owned IP assets which were running "restconfig" services hosted on Swagger-UI. Dawid Moczadło has written a detailed article here . POC - Three IP assets on -  "https://IP/ui" were vulnerable to DOM XSS on this endpoint - "/ui/model.html?configUrl=http://hackingmonks/test.json". The Swagger-UI did show the available API endpoints for the service but had authorisation in place(access_token was needed). During this period, I had found 4th asset which had the same vulnerability along with additional exploitation, which was reported in another ticket. The write-up can be found here ( Access to BGP server + DOM XSS ). Timeline - Reported - 22.05.2023 Triaged - 22.05.2023 The assets were no longer accessible to the public -  09.06.2023 Accepted - 07.06.2023 Rewarded - 13.06.2023 Google's date for the fix - 08.12.2023
Read more

[Google] Disclose hidden Blogger profile Display name and Profile photo

 Because of an unfiltered response to an API call, it was possible for anyone to fetch hidden Blogger Profile Display name and Profile photo. POC - An endpoint going to - ========================================= GET https://blogger.googleapis.com/v3/blogs/target_blog_id/posts?fetchBodies=true&fetchImages=true&maxResults=69&view=VIEW_TYPE_UNSPECIFIED&key=[YOUR_API_KEY] HTTP/1.1 Authorization: Bearer [YOUR_ACCESS_TOKEN] ========================================= Would respond with the hidden Profile Display name and Profile photo as below - ========================================= "author": {     "id": "blogger_id",     "displayName": "blogger_name",     "url": "https://www.blogger.com/profile/blogger_id",     "image": {       "url": "//2.bp.blogspot.com/blogger_photo.jpg" } ========================================= Timeline - Reported - 31.01.2023 Triaged - 01.02.2023 Acc...
Read more