Dark Reading is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Application Security

3/20/2019
02:30 PM
Connect Directly
Twitter
LinkedIn
Google+
RSS
E-Mail
100%
0%

Google Photos Bug Let Criminals Query Friends, Location

The vulnerability, now patched, let attackers query where, when, and with whom victims' photos were taken.

A now-patched vulnerability in the Web version of Google Photos could let cybercriminals learn the details of a user's photo history, Imperva reports.

Through browser-based timing attacks, hackers could analyze image data to learn when a person visited a particular place. It's not a common threat, and it's most effective in a targeted scenario, but it was possible for someone to use a malicious website to access photo data.

Google Photos knows a lot about the people who use it. The service automatically tags each image using metadata (date, location coordinates), and its artificial intelligence engine can detect objects and events that would indicate a wedding, waterfall, sunset, or range of other locations, explains Imperva researcher Ron Masas. Facial recognition tags people also present in the photos.

Coupled with Google Photos' powerful search engine, this detailed information could share a lot about when, where, and with whom a person has been. All of this data can be used in search queries to unearth certain photos – for example: "Photos of me and Ashley from Paris 2018."

Masas decided to investigate Google Photos for side-channel attacks when he learned the extent of its search capabilities. He found the service's search endpoint is vulnerable to an attack called Cross-Site Search, or XS-Search. In a proof of concept, he used the HTML link tag to create multiple cross-origin requests to the Photos search endpoint. Using JavaScript, he measured the time it took for the onload event to trigger. This is the baseline time, or the time it took to query Google Photos' server and receive zero photo results as a response.

With this baseline time, Masas queried "photos of me from Iceland" and compared the two times. If this search took longer, he could infer the user had visited Iceland based on the data. If he added a date, he could know whether photos were taken in a specific time frame. For every place queried, a time longer than the baseline time indicates the user took photos there.

Here's how this vulnerability works: An attacker would have to first send the target a malicious link while that person is logged into Google Photos, by embedding malicious JavaScript inside a Web advertisement or sending a direct message via email or a online messaging service. Malicious JavaScript code creates requests to the Google Photos search endpoint and extracts answers.

"The vulnerability is basically allowing different sites to search for you," Masas explains. As the malicious page is open, an attacker could repeatedly query Google Photos in the background. "By using the advanced search feature, I can ask a lot of questions about you," he adds.

However, once the victim closes the malicious page, the searches stop. "The moment you close the site, I no longer can do that," Masas says. "But I can trick you into opening another site in the future and can continue from there. It does require you to open a website each time."

In his opinion, this isn't a very complex attack but it does have the most value if a hacker is specifically targeting one individual. For example, someone could have used this to determine the location of a high-profile person or know who they have been spending time with. This type of attack "is very hard to detect if you're not looking for it actively," Masas adds. Could a similar vulnerability exist in other online services and applications? "Definitely," he notes. Many developers aren't aware of this problem, and it's important large and small sites learn of it.

This isn't the first time Masas has exposed an attack of this nature. Earlier this month, Facebook patched a flaw he had discovered in the Web version of Facebook Messenger, which could have let attackers view the people with whom someone had been chatting. Similarly, a victim would have to be tricked into opening a bad link; he'd also have to click somewhere on the page. When a new tab opened, the attacker could use the previous page to load the messenger chat endpoint and view which specific person or bots the target had been talking to.

Related Content:

 

 

Join Dark Reading LIVE for two cybersecurity summits at Interop 2019. Learn from the industry's most knowledgeable IT security experts. Check out the Interop agenda here.

Kelly Sheridan is the Staff Editor at Dark Reading, where she focuses on cybersecurity news and analysis. She is a business technology journalist who previously reported for InformationWeek, where she covered Microsoft, and Insurance & Technology, where she covered financial ... View Full Bio

Comment  | 
Print  | 
More Insights
Comments
Newest First  |  Oldest First  |  Threaded View
AI Is Everywhere, but Don't Ignore the Basics
Howie Xu, Vice President of AI and Machine Learning at Zscaler,  9/10/2019
Fed Kaspersky Ban Made Permanent by New Rules
Dark Reading Staff 9/11/2019
Register for Dark Reading Newsletters
White Papers
Video
Cartoon Contest
Current Issue
7 Threats & Disruptive Forces Changing the Face of Cybersecurity
This Dark Reading Tech Digest gives an in-depth look at the biggest emerging threats and disruptive forces that are changing the face of cybersecurity today.
Flash Poll
Twitter Feed
Dark Reading - Bug Report
Bug Report
Enterprise Vulnerabilities
From DHS/US-CERT's National Vulnerability Database
CVE-2019-16317
PUBLISHED: 2019-09-14
In Pimcore before 5.7.1, an attacker with limited privileges can trigger execution of a .phar file via a phar:// URL in a filename parameter, because PHAR uploads are not blocked and are reachable within the phar://../../../../../../../../var/www/html/web/var/assets/ directory, a different vulnerabi...
CVE-2019-16318
PUBLISHED: 2019-09-14
In Pimcore before 5.7.1, an attacker with limited privileges can bypass file-extension restrictions via a 256-character filename, as demonstrated by the failure of automatic renaming of .php to .php.txt for long filenames, a different vulnerability than CVE-2019-10867 and CVE-2019-16317.
CVE-2019-16307
PUBLISHED: 2019-09-14
A Reflected Cross-Site Scripting (XSS) vulnerability in the webEx module in webExMeetingLogin.jsp and deleteWebExMeetingCheck.jsp in Fuji Xerox DocuShare through 7.0.0.C1.609 allows remote attackers to inject arbitrary web script or HTML via the handle parameter (webExMeetingLogin.jsp) and meetingKe...
CVE-2019-16294
PUBLISHED: 2019-09-14
SciLexer.dll in Scintilla in Notepad++ (x64) before 7.7 allows remote code execution or denial of service via Unicode characters in a crafted .ml file.
CVE-2019-16309
PUBLISHED: 2019-09-14
FlameCMS 3.3.5 has SQL injection in account/login.php via accountName.