Researchers Show Spectre Can Still Bite in Cloudflare Workers

Cybersecurity researchers have disclosed details of a remote Spectre attack against Cloudflare Workers that leaked a JSON Web Token (JWT) from a co-located Worker in the production environment at up to 12 bits per second. That’s 360 times the rate of an earlier attack demonstrated in 2021.

The end-to-end experiment used an attacker Worker and a victim Worker controlled by the researchers, both running on the same physical machine. The attack exploited speculative execution to leak the JWT — a credential used for authentication — in real time.

The findings, presented at a recent security conference, underscore that Spectre-class vulnerabilities remain a practical threat even in managed serverless platforms. The researchers achieved a leak rate of 12 bits per second, which might sound slow, but it’s enough to exfiltrate a typical JWT in under a minute.

How the Attack Works

The attack relies on the same fundamental flaw as the original Spectre: speculative execution. By tricking the CPU into executing instructions out of order, the attacker can access memory that should be off-limits.

In this case, the attacker Worker uses a timing side channel to read sensitive data from the victim Worker’s memory. The researchers demonstrated that the technique works reliably in Cloudflare’s production environment, not just in a lab.

The 2021 Baseline

Back in 2021, a similar attack was shown to leak data at just 0.033 bits per second. That was largely theoretical — too slow to be practical. The new attack’s 12 bits per second changes the calculus.

“At this rate, stealing a full JWT is not just possible but feasible within a short window,” the researchers noted in their paper.

Why This Matters for Cloudflare Users

Cloudflare Workers is a popular serverless platform used by millions of developers. If you’re running Workers that handle sensitive data — like JWTs, API keys, or user PII — this attack is a wake-up call.

The researchers emphasized that the attack requires co-location, meaning the attacker must land on the same physical server as the victim. Cloudflare’s isolation measures make this harder, but not impossible.

For developers, the practical takeaway is to avoid storing long-lived secrets in Worker memory. Rotate JWTs frequently and use short expiry times to limit the damage if a leak occurs.

Cloudflare’s Response and Mitigations

Cloudflare has acknowledged the research and implemented additional mitigations. The company has a history of hardening its platform against Spectre, but this new attack shows that the arms race is ongoing.

Mitigations include:

• Enhanced isolation between Workers on the same machine
• Randomized memory layouts to make timing attacks harder
• Ongoing monitoring for speculative execution anomalies

Cloudflare also advises users to follow best practices for secret management, such as using environment variables and avoiding hardcoded credentials.

Broader Implications for Serverless Security

This research isn’t just about Cloudflare. It highlights a broader issue: serverless platforms are not immune to CPU-level attacks. Any platform that runs untrusted code on shared hardware is potentially vulnerable.

Other providers like AWS Lambda and Google Cloud Functions face similar challenges, though they haven’t yet seen public demonstrations at this speed.

The researchers suggest that hardware-level fixes, like those in newer CPUs, are the long-term solution. Until then, software mitigations and vigilant secret management are the best defense.

What Developers Should Do Now

If you’re a Cloudflare Workers user, don’t panic — but do take action. Start by auditing your Workers for sensitive data in memory. Use short-lived JWTs and rotate them frequently.

Consider using Cloudflare’s built-in secrets management and avoid storing tokens in global variables. Also, keep an eye on Cloudflare’s security advisories for updates.

The 12 bits per second leak rate is a proof of concept, but it’s a serious one. As the researchers put it, “The era of dismissing Spectre as theoretical is over.”

For more on related topics, check out our coverage of Spectre mitigation strategies and serverless security best practices.

Citrix NetScaler Authentication Bypass: The 9.3 Severity Flaw

Citrix has rolled out patches for two vulnerabilities in its NetScaler ADC and NetScaler Gateway products. The more serious of the two, tracked as CVE-2026-19490, carries a CVSS score of 9.3 and is described as an authentication bypass using an alternative path.

This Citrix NetScaler authentication bypass impacts appliances configured as a gateway (SSL VPN, ICA Proxy, CVPN, RDP Proxy) or an AAA virtual server. The worst part? It can be exploited by remote, unauthenticated attackers without any user interaction, according to cybersecurity firm Rapid7.

Which NetScaler Versions Are Vulnerable?

Per Citrix’s advisory, the security defect impacts a wide range of NetScaler ADC and NetScaler Gateway versions. The list includes 14.1-43.56 or later, 14.1-66.68-FIPS or later, 14.1-43.55 or earlier, 13.1-61.28 or later, 13.1-61.27 or earlier, and 13.1 FIPS.

If you’re running any of these, you need to act. The fixed versions are:

• 14.1-73.32
• 13.1-63.21
• 14.1-73.32 FIPS
• 13.1-FIPS and 13.1-NDcPP 13.1-37.277

These builds also contain fixes for CVE-2026-19489, a high-severity memory overflow issue. That one could lead to unexpected behavior or denial-of-service (DoS) if SIP ALG is enabled at an LSN group configuration.

Secure Private Access Hybrid Deployments Also Affected

Citrix notes that Secure Private Access Hybrid deployments using NetScaler instances are also affected by these vulnerabilities. Customers need to upgrade these NetScaler instances to the recommended builds to address the issues.

Why Experts Expect Quick Exploitation

Rapid7 says there are currently no indicators that threat actors are exploiting the authentication bypass issue. But that’s likely to change. NetScaler’s critical role within enterprise systems makes it an attractive target for hackers.

The firm points out that NetScaler ADC and NetScaler Gateway are widely deployed enterprise networking products commonly positioned at or near the network perimeter. NetScaler ADC provides application delivery, traffic management, load balancing, SSL/TLS offloading, and application security capabilities, while NetScaler Gateway provides secure remote access and VPN functionality.

Given that NetScaler appliances are typically deployed in enterprise DMZs and are publicly accessible, Rapid7 expects threat actors to exploit the critical bug shortly. The company’s advice is blunt: organizations should prioritize patching affected systems on an emergency basis, since Citrix products are high-value targets that tend to quickly see exploitation in the wild.

This isn’t the first time Citrix NetScaler has been in the spotlight for security issues. Past incidents have shown that attackers move fast when it comes to these devices. For a broader look at how similar flaws have played out, check out our coverage of the critical GitLab flaw exploited shortly after disclosure and the previous Citrix NetScaler HTTP/2 bomb attack.

What Should Organizations Do Now?

The message from security researchers is clear: don’t wait. If your organization uses Citrix NetScaler ADC or Gateway, verify your version and apply the patches immediately. The authentication bypass issue is too dangerous to leave unaddressed, especially given the public accessibility of these devices.

For those who need more context on how similar vulnerabilities have been handled, our article on the exploitation of a fresh Citrix NetScaler vulnerability offers useful background on the threat landscape surrounding these products.

Stay ahead of the curve by keeping your NetScaler deployments up to date and monitoring Citrix’s security advisories regularly. The window for safe patching may be smaller than you think.

The Discovery of SilkParasite

Late 2025 brought a quiet but significant shift in the cyber threat landscape. Researchers uncovered an intrusion set, now dubbed SilkParasite, that had been quietly burrowing into government networks across Central Asia. The operation wasn’t flashy. It didn’t scream for attention. Instead, it relied on a toolkit of seven distinct remote access trojans (RATs), five of which had never been seen in the wild before.

Security analysts first flagged the cluster in November 2025, after detecting anomalous traffic patterns from a handful of government ministries. The investigation revealed a sophisticated, multi-stage operation that had likely been active for months, possibly longer. The five new tools — DriveSilkRAT, CookiETagRAT, NomadRAT, GoginRAT, and NodeEdgeRAT — represent a worrying evolution in modular malware design.

The SilkParasite espionage campaign stands out not just for its arsenal, but for its patience. This wasn’t a smash-and-grab. It was a slow, deliberate operation designed to maintain persistent access to sensitive government systems.

Five New RATs, One Cohesive Arsenal

Each of the five newly discovered RATs serves a specific function, a modular approach that allows the operators to swap tools in and out depending on the target environment. Here’s a breakdown of what makes each one distinct:

• DriveSilkRAT — Focuses on file system manipulation and data exfiltration from connected drives.
• CookiETagRAT — Uses HTTP cookie headers as a covert command-and-control channel, blending traffic with normal web browsing.
• NomadRAT — A lightweight, evasive tool designed for reconnaissance and lateral movement.
• GoginRAT — Written in Go, this RAT offers cross-platform capabilities and robust encryption for its communications.
• NodeEdgeRAT — Exploits edge device vulnerabilities to gain initial footholds in network perimeters.

This diversity suggests a well-funded operation with significant development resources. The use of new RATs malware in tandem with two previously known tools indicates a group that’s constantly iterating on its tradecraft.

Why the Modular Design Matters

For defenders, the modular structure is a nightmare. Traditional signature-based detection fails against tools that can be reconfigured on the fly. If one RAT is burned, the operators simply switch to another. This redundancy is a hallmark of advanced persistent threat (APT) groups, and it’s what makes the Central Asian government cyberattack so difficult to fully remediate.

Targeting Government Networks in Central Asia

The campaign’s focus on government bodies is a clear indicator of espionage intent. While the researchers didn’t name specific agencies, the geographic concentration points to a strategic interest in the region’s political, economic, and energy infrastructure.

Central Asian governments have become increasingly attractive targets for cyber espionage in recent years. Their growing digital infrastructure, combined with geopolitical tensions, creates a perfect storm. The attackers likely used spear-phishing emails and zero-day exploits in edge devices to gain initial access, then deployed the RATs to establish persistence.

One particularly cunning technique observed involves CookiETagRAT abusing HTTP cookie headers. By hiding commands in what looks like standard web traffic, the malware can evade many network monitoring tools. It’s a clever trick, and it’s one that highlights the increasing sophistication of the cyber espionage campaign.

Defending Against an Adaptive Adversary

So, what should security teams in the region — and elsewhere — take away from this? First, assume your perimeter is already compromised. The days of relying solely on firewalls and antivirus are over. Second, invest in behavior-based detection. The SilkParasite RATs rely on mimicking normal traffic patterns, so anomaly detection is key.

Third, patch edge devices aggressively. NodeEdgeRAT specifically targets vulnerabilities in routers and VPNs. If those aren’t patched, you’re leaving the front door open.

Finally, segment your networks. If an attacker gains access to one system, segmentation limits their ability to move laterally. The SilkParasite campaign shows that remote access trojans are becoming more specialized and more evasive. Staying ahead requires a proactive, layered defense strategy.

The discovery of SilkParasite is a reminder that cyber espionage isn’t slowing down. It’s getting quieter, smarter, and more dangerous. For governments in Central Asia, the message is clear: the parasites are already inside. The only question is whether you can root them out before they achieve their mission.