Entries Tagged "browsers"

Page 1 of 7

Prompt Injection in AI Browsers

This is why AIs are not ready to be personal assistants:

A new attack called ‘CometJacking’ exploits URL parameters to pass to Perplexity’s Comet AI browser hidden instructions that allow access to sensitive data from connected services, like email and calendar.

In a realistic scenario, no credentials or user interaction are required and a threat actor can leverage the attack by simply exposing a maliciously crafted URL to targeted users.

[…]

CometJacking is a prompt-injection attack where the query string processed by the Comet AI browser contains malicious instructions added using the ‘collection’ parameter of the URL.

LayerX researchers say that the prompt tells the agent to consult its memory and connected services instead of searching the web. As the AI tool is connected to various services, an attacker leveraging the CometJacking method could exfiltrate available data.

In their tests, the connected services and accessible data include Google Calendar invites and Gmail messages and the malicious prompt included instructions to encode the sensitive data in base64 and then exfiltrate them to an external endpoint.

According to the researchers, Comet followed the instructions and delivered the information to an external system controlled by the attacker, evading Perplexity’s checks.

I wrote previously:

Prompt injection isn’t just a minor security problem we need to deal with. It’s a fundamental property of current LLM technology. The systems have no ability to separate trusted commands from untrusted data, and there are an infinite number of prompt injection attacks with no way to block them as a class. We need some new fundamental science of LLMs before we can solve this.

Posted on November 11, 2025 at 7:08 AM • View Comments

New Windows Malware Locks Computer in Kiosk Mode

Clever:

A malware campaign uses the unusual method of locking users in their browser’s kiosk mode to annoy them into entering their Google credentials, which are then stolen by information-stealing malware.

Specifically, the malware “locks” the user’s browser on Google’s login page with no obvious way to close the window, as the malware also blocks the “ESC” and “F11” keyboard keys. The goal is to frustrate the user enough that they enter and save their Google credentials in the browser to “unlock” the computer.

Once credentials are saved, the StealC information-stealing malware steals them from the credential store and sends them back to the attacker.

I’m sure this works often enough to be a useful ploy.

Posted on September 25, 2024 at 7:00 AM • View Comments

Exploiting Mistyped URLs

Interesting research: “Hyperlink Hijacking: Exploiting Erroneous URL Links to Phantom Domains“:

Abstract: Web users often follow hyperlinks hastily, expecting them to be correctly programmed. However, it is possible those links contain typos or other mistakes. By discovering active but erroneous hyperlinks, a malicious actor can spoof a website or service, impersonating the expected content and phishing private information. In “typosquatting,” misspellings of common domains are registered to exploit errors when users mistype a web address. Yet, no prior research has been dedicated to situations where the linking errors of web publishers (i.e. developers and content contributors) propagate to users. We hypothesize that these “hijackable hyperlinks” exist in large quantities with the potential to generate substantial traffic. Analyzing large-scale crawls of the web using high-performance computing, we show the web currently contains active links to more than 572,000 dot-com domains that have never been registered, what we term ‘phantom domains.’ Registering 51 of these, we see 88% of phantom domains exceeding the traffic of a control domain, with up to 10 times more visits. Our analysis shows that these links exist due to 17 common publisher error modes, with the phantom domains they point to free for anyone to purchase and exploit for under $20, representing a low barrier to entry for potential attackers.

Posted on June 10, 2024 at 7:08 AM • View Comments

Class-Action Lawsuit against Google’s Incognito Mode

The lawsuit has been settled:

Google has agreed to delete “billions of data records” the company collected while users browsed the web using Incognito mode, according to documents filed in federal court in San Francisco on Monday. The agreement, part of a settlement in a class action lawsuit filed in 2020, caps off years of disclosures about Google’s practices that shed light on how much data the tech giant siphons from its users­—even when they’re in private-browsing mode.

Under the terms of the settlement, Google must further update the Incognito mode “splash page” that appears anytime you open an Incognito mode Chrome window after previously updating it in January. The Incognito splash page will explicitly state that Google collects data from third-party websites “regardless of which browsing or browser mode you use,” and stipulate that “third-party sites and apps that integrate our services may still share information with Google,” among other changes. Details about Google’s private-browsing data collection must also appear in the company’s privacy policy.

I was an expert witness for the prosecution (that’s the class, against Google). I don’t know if my declarations and deposition will become public.

Posted on April 3, 2024 at 7:01 AM • View Comments

An Untrustworthy TLS Certificate in Browsers

The major browsers natively trust a whole bunch of certificate authorities, and some of them are really sketchy:

Google’s Chrome, Apple’s Safari, nonprofit Firefox and others allow the company, TrustCor Systems, to act as what’s known as a root certificate authority, a powerful spot in the internet’s infrastructure that guarantees websites are not fake, guiding users to them seamlessly.

The company’s Panamanian registration records show that it has the identical slate of officers, agents and partners as a spyware maker identified this year as an affiliate of Arizona-based Packet Forensics, which public contracting records and company documents show has sold communication interception services to U.S. government agencies for more than a decade.

[…]

In the earlier spyware matter, researchers Joel Reardon of the University of Calgary and Serge Egelman of the University of California at Berkeley found that a Panamanian company, Measurement Systems, had been paying developers to include code in a variety of innocuous apps to record and transmit users’ phone numbers, email addresses and exact locations. They estimated that those apps were downloaded more than 60 million times, including 10 million downloads of Muslim prayer apps.

Measurement Systems’ website was registered by Vostrom Holdings, according to historic domain name records. Vostrom filed papers in 2007 to do business as Packet Forensics, according to Virginia state records. Measurement Systems was registered in Virginia by Saulino, according to another state filing.

More details by Reardon.

Cory Doctorow does a great job explaining the context and the general security issues.

EDITED TO ADD (11/10): Slashdot thread.

Posted on November 10, 2022 at 9:18 AM • View Comments

Leaking Passwords through the Spellchecker

Sometimes browser spellcheckers leak passwords:

When using major web browsers like Chrome and Edge, your form data is transmitted to Google and Microsoft, respectively, should enhanced spellcheck features be enabled.

Depending on the website you visit, the form data may itself include PII­—including but not limited to Social Security Numbers (SSNs)/Social Insurance Numbers (SINs), name, address, email, date of birth (DOB), contact information, bank and payment information, and so on.

The solution is to only use the spellchecker options that keep the data on your computer—and don’t send it into the cloud.

Posted on September 26, 2022 at 6:08 AM • View Comments

Facebook Is Now Encrypting Links to Prevent URL Stripping

Some sites, including Facebook, add parameters to the web address for tracking purposes. These parameters have no functionality that is relevant to the user, but sites rely on them to track users across pages and properties.

Mozilla introduced support for URL stripping in Firefox 102, which it launched in June 2022. Firefox removes tracking parameters from web addresses automatically, but only in private browsing mode or when the browser’s Tracking Protection feature is set to strict. Firefox users may enable URL stripping in all Firefox modes, but this requires manual configuration. Brave Browser strips known tracking parameters from web addresses as well.

Facebook has responded by encrypting the entire URL into a single ciphertext blob.

Since it is no longer possible to identify the tracking part of the web address, it is no longer possible to remove it from the address automatically. In other words: Facebook has the upper hand in regards to URL-based tracking at the time, and there is little that can be done about it short of finding a way to decrypt the information.

Posted on July 18, 2022 at 9:49 AM • View Comments

New Browser De-anonymization Technique

Researchers have a new way to de-anonymize browser users, by correlating their behavior on one account with their behavior on another:

The findings, which NJIT researchers will present at the Usenix Security Symposium in Boston next month, show how an attacker who tricks someone into loading a malicious website can determine whether that visitor controls a particular public identifier, like an email address or social media account, thus linking the visitor to a piece of potentially personal data.

When you visit a website, the page can capture your IP address, but this doesn’t necessarily give the site owner enough information to individually identify you. Instead, the hack analyzes subtle features of a potential target’s browser activity to determine whether they are logged into an account for an array of services, from YouTube and Dropbox to Twitter, Facebook, TikTok, and more. Plus the attacks work against every major browser, including the anonymity-focused Tor Browser.

[…]

“Let’s say you have a forum for underground extremists or activists, and a law enforcement agency has covertly taken control of it,” Curtmola says. “They want to identify the users of this forum but can’t do this directly because the users use pseudonyms. But let’s say that the agency was able to also gather a list of Facebook accounts who are suspected to be users of this forum. They would now be able to correlate whoever visits the forum with a specific Facebook identity.”

Posted on July 14, 2022 at 9:31 AM • View Comments

Is Microsoft Stealing People’s Bookmarks?

I received email from two people who told me that Microsoft Edge enabled synching without warning or consent, which means that Microsoft sucked up all of their bookmarks. Of course they can turn synching off, but it’s too late.

Has this happened to anyone else, or was this user error of some sort? If this is real, can some reporter write about it?

(Not that “user error” is a good justification. Any system where making a simple mistake means that you’ve forever lost your privacy isn’t a good one. We see this same situation with sharing contact lists with apps on smartphones. Apps will repeatedly ask, and only need you to accidentally click “okay” once.)

EDITED TO ADD: It’s actually worse than I thought. Edge urges users to store passwords, ID numbers, and even passport numbers, all of which get uploaded to Microsoft by default when synch is enabled.

Posted on November 17, 2021 at 7:53 AM • View Comments

1 2 3 … 7

Sidebar photo of Bruce Schneier by Joe MacInnis.