Malicious Perplexity Chrome Extension Intercepted Searches and Address Bar Input

# Fake Perplexity Chrome Extension Silently Logged Searches and Keystrokes—Here's What We Know

A malicious Chrome extension impersonating the popular AI search engine Perplexity intercepted user searches and address bar input for an unknown period before Microsoft researchers discovered and reported it. The extension, which has since been removed from the Chrome Web Store by Google following responsible disclosure, represents a sophisticated example of how supply-chain attacks and search-adjacent targets continue to evolve in the threat landscape.

## The Threat: Man-in-the-Middle for Every Query

The malicious extension operated as a transparent intermediary between users and their browsing activity. Rather than blocking or replacing legitimate Perplexity functionality, it monitored everything users typed—both search queries and address bar input—and transmitted this data to an attacker-controlled server before allowing the user's requests to proceed normally.

This approach is particularly dangerous because it maintains the illusion of normal operation. Users saw results they expected, faced no obvious errors, and had no clear indication that their keystrokes were being harvested. From a user perspective, the extension appeared to work exactly as intended.

Key details about the attack:

## Background and Context: Why Perplexity?

Perplexity AI, launched in 2023, has become one of the fastest-growing search and research tools, particularly among technology professionals, researchers, and knowledge workers. The platform uses large language models to synthesize web results and provide conversational search capabilities—a direct alternative to both traditional Google Search and newer AI chatbots.

The extension ecosystem represents an ongoing security challenge. Extensions have broad permissions by default, including the ability to monitor and modify network traffic, access browsing history, and intercept keystrokes. While legitimate extensions require these permissions to function, malicious actors exploit the same APIs for surveillance and data theft.

Why extensions remain attractive targets:

## Technical Details: How the Attack Worked

The extension functioned as a proxy, intercepting HTTP requests before they reached their destination. When a user typed a query or clicked search, the extension captured the request, forwarded it to an attacker-controlled server for logging, then allowed the request to proceed to Perplexity's legitimate infrastructure.

Attack flow:

1. User installs what appears to be official Perplexity extension from Chrome Web Store

2. User performs search or types in address bar

3. Extension intercepts the request

4. Data is logged on attacker server (search term, timestamp, user context)

5. Request is forwarded to real Perplexity servers

6. User receives normal search results

This "pass-through" design is a hallmark of sophisticated credential and data-harvesting attacks. By not disrupting legitimate functionality, the attacker minimizes the risk of detection. Users would have no reason to suspect the extension was malicious unless they:

Microsoft's detection likely came through either threat intelligence monitoring of suspicious extension behavior or discovery of infrastructure associated with the attacker's command-and-control servers.

## Who Was Targeted?

The scope of compromise remains unclear. Google's extension metrics do not always reflect download counts, and Microsoft's disclosure did not specify:

However, given that Perplexity is particularly popular among technology professionals, security researchers, and knowledge workers, the attacker likely harvested intelligence valuable to multiple threat actors—search behavior correlates with research interests, job hunting (visible in search patterns), financial decisions, and security work.

## Implications: A Broader Pattern

This incident exemplifies several converging trends in the threat landscape:

1. Search as a high-value target

Search queries reveal intent, interests, vulnerabilities, and decision-making. An attacker who can log searches gains intelligence on competitor research, security investigations, acquisition targets, and personal financial planning. This data is valuable both for direct social engineering and for resale to other threat actors.

2. Extension-based supply chain attacks

Extensions distribute with the implicit trust of the Chrome Web Store. Users assume Google's review process screens for malicious behavior, but this assumption has been repeatedly disproven. Attackers can create seemingly legitimate extensions, accumulate users, and activate malicious behavior later—or operate covertly from day one.

3. Ease of infrastructure setup

Setting up an attacker-controlled logging server requires minimal resources. The attacker needed only a domain, hosting, and a few lines of code to capture and store keystroke data. This low barrier to entry means capability does not indicate sophisticated actors—this attack could originate from individual criminals, corporate espionage teams, or nation-state threat actors.

## Recommendations for Users and Organizations

For individual users:

For organizations:

## HackWire Analysis

This attack reflects a fundamental security asymmetry: users trust the official extension marketplaces more than they scrutinize what those extensions actually do. Google's review process catches obvious malware, but sophisticated pass-through proxies—which maintain perfect compatibility with legitimate use—can slip through.

What makes this incident particularly noteworthy is the targeting of search behavior itself. Unlike credential-stealing malware or banking trojans that focus on specific account types, keystroke logging on search gives an attacker a lens into the victim's thinking process. An attacker who monitors searches can identify which companies a person is researching, which security tools they're evaluating, what health conditions they're investigating, and what job opportunities they're exploring. This intelligence is valuable for targeted spear-phishing, social engineering, and competitive intelligence.

The Perplexity extension specifically is a clever choice because it targets the segment of users most likely to be information workers, researchers, and security professionals—exactly the people whose search behavior and research interests are most valuable. A attacker harvesting from Perplexity users gains insight into who is investigating what, who is asking security questions, and who is researching emerging threats.

The incident also underscores why users should prefer web-based tools over browser extensions whenever possible. Perplexity, like most modern AI search tools, operates perfectly well as a web application. An extension adds convenience but introduces a layer of privilege that malicious actors can exploit. The security trade-off is rarely worth it.

For defenders, this serves as a reminder: extension-based attacks are not edge cases. They're practical, scalable, and difficult to detect without active monitoring. Organizations should treat extension management as a core security control, not an afterthought.

— HackWire Editorial

## Recommendations Going Forward

Organizations should use this incident as a trigger for deeper security hygiene:

1. Conduct an extension inventory across all endpoints and remove anything non-essential

2. Deploy detection rules for unusual outbound traffic from browser processes

3. Educate users on the risks of convenience tools and the value of their search behavior

4. Monitor threat feeds for similar fake-AI-tool extensions that follow this pattern

5. Test user awareness with simulated extension installation scenarios

---

## Related Coverage