China-State IoT Botnet Functioned Undetected for Four Years, Recently Uncovered

# FBI Disassembles Vast Chinese State-Sponsored Botnet: Raptor Train

In a notable achievement for cybersecurity, the FBI has effectively dismantled a vast botnet referred to as **Raptor Train**, which had been utilized by Chinese state-sponsored hackers for more than four years. This botnet aimed at a diverse array of entities, including government bodies, telecommunications firms, defense contractors, and other essential infrastructure in the United States and Taiwan. This operation represents a significant advancement in the persistent conflict against state-sponsored cyber intrusions, especially those emanating from China.

## The Extent of the Botnet

Raptor Train was mainly constituted of compromised **Internet of Things (IoT)** devices, such as small office and home routers, surveillance cameras, and network-attached storage (NAS) devices. Over the past four years, over **260,000 devices** cycled through the botnet, with **60,000 devices** active at its zenith in June 2023, as per researchers from **Black Lotus Labs**. This categorizes Raptor Train as the largest botnet connected to the Chinese state ever uncovered.

The structure of the botnet was organized into a **three-tier framework** that enabled it to function with exceptional efficiency and accuracy. The affected devices were distributed worldwide, with more than half situated in North America and another 25% in Europe.

## A Second Major Seizure in 2024

Raptor Train is the second botnet associated with the Chinese state that the FBI has disassembled in 2024. Earlier this year, the FBI took down another botnet linked to a Chinese hacking faction named **Volt Typhoon**. Similar to Raptor Train, Volt Typhoon leveraged compromised IoT devices to initiate assaults on vital infrastructure. These botnets pose a particular threat as they utilize IP addresses from trusted devices, enabling them to circumvent numerous network security measures.

The **Chinese hackers** responsible for Raptor Train are affiliated with a group identified as **Flax Typhoon**, which has connections to **Integrity Technology Group**, a China-based entity associated with the **People’s Republic of China**. The company deployed state-controlled IP addresses from **China Unicom Beijing Province Network** to oversee the botnet.

## FBI’s Action to Dismantle Raptor Train

The FBI’s campaign against Raptor Train consisted of **court-sanctioned operations** that permitted law enforcement to seize control of the botnet’s infrastructure. This involved identifying compromised devices and issuing covert commands to cleanse them from the malware. FBI Director **Christopher Wray** stated that when the hackers acknowledged that their botnet was being dismantled, they sought to migrate their bots to new servers and even executed a **Distributed Denial of Service (DDoS)** attack against the FBI. Nevertheless, the FBI and its allies swiftly managed the attack and pinpointed the new infrastructure, compelling the hackers to relinquish their botnet.

## The Framework of Raptor Train

Raptor Train’s intricate structure facilitated its operation on a vast scale. The botnet was structured into three tiers:

1. **Tier 1**: This tier was composed of compromised IoT devices like routers, IP cameras, and NAS devices. These gadgets were infected with a specific malware known as **Nosedive**, which is based on **Mirai**, a infamous IoT malware responsible for unprecedented DDoS attacks in 2016. Many of the compromised devices were at the end of their operational life, meaning they no longer received security updates, rendering them easy prey for exploitation. The FBI identified more than **70 vulnerabilities** that were exploited to compromise these devices from 2019 to 2024.

2. **Tier 2**: This tier consisted of virtual servers that acted as **command-and-control (C2)** servers for the Tier 1 devices. These servers transmitted exploits and payloads to the infected devices and maintained communication using a distinctive URL encoding method and domain injection technique. The number of Tier 2 nodes steadily increased over the years, reaching **60 nodes** by August 2024.

3. **Tier 3**: This tier included a limited number of **management nodes** operating specialized software named **Sparrow**. These nodes enabled the Flax Typhoon hackers to manually control the Tier 2 nodes and gather intelligence from the botnet. Interactions between Tier 2 and Tier 3 typically occurred during Chinese business hours, signifying the participation of state-sponsored actors.

## Attacking Critical Infrastructure

The botnet was primarily intended to target critical infrastructure in the US and Taiwan, encompassing **military, governmental, telecommunications, defense industrial base (DIB)**, and **higher education** sectors. In late 2023, the botnet operators engaged in extensive scanning operations aimed at US military and governmental systems, in addition to IT service providers and DIBs. They also targeted vulnerable software, such as **Atlassian Confluence** servers and **Ivanti Connect Secure** appliances, leveraging known

Lionsgate Collaborates in Major Agreement to Provide Film Library for AI Training

**AI and Hollywood: The Lionsgate-Runway Alliance and Its Impact on Future Content Production**

On Wednesday, the AI video synthesis firm Runway and the entertainment powerhouse Lionsgate unveiled a revolutionary partnership intended to transform the production of films and television series. This collaboration will involve Runway creating a specialized AI model trained on Lionsgate’s vast library of films and TV shows, aiming to enhance content production while potentially lowering costs. This agreement marks a crucial move towards the incorporation of artificial intelligence into Hollywood’s creative workflows, generating both enthusiasm and apprehension within the sector.

### The Alliance: A New Chapter in AI-Enhanced Content Production

Lionsgate, recognized for hit franchises like *John Wick* and *The Hunger Games*, is turning to AI as a means to optimize production and improve efficiency. Michael Burns, Lionsgate’s vice chair, noted that AI could facilitate the development of “state-of-the-art, cost-effective content generation opportunities.” He also mentioned that various filmmakers are already showing interest in the prospective uses of AI during pre- and post-production stages.

Runway, a pioneer in AI-based video synthesis, intends to build a unique AI model that will be exclusive to Lionsgate Studios. This model will empower filmmakers, directors, and creative teams to enhance their projects with AI-generated elements. While the precise applications of this model remain unspecified, this partnership represents the first major alliance between an AI video synthesis company and a Hollywood studio, signaling a potential shift in the entertainment industry’s content creation methods.

Cristóbal Valenzuela, co-founder and CEO of Runway, highlighted the transformative capabilities of this technology: “We’re dedicated to providing artists, creators, and studios with the most advanced and effective tools to enhance their workflows and explore new methods of storytelling. The evolution of art is intertwined with technology, and these innovative models are part of our ongoing mission to create transformative mediums for artistic and creative expression; the most compelling stories are yet to come.”

### The Search for Licensed Training Data

AI generative models, such as those produced by Runway, depend on substantial amounts of data to operate successfully. These models scrutinize existing video and image examples to identify patterns that can then be utilized to create new content based on textual prompts. However, the procurement of this training data poses legal challenges, especially when it pertains to copyrighted materials.

Runway currently faces a class-action lawsuit alleging that the company utilized video data acquired without authorization to train its AI models, raising issues of copyright violation. While some companies, including OpenAI, contend that collecting copyrighted content for AI training qualifies as “fair use,” U.S. courts have not yet definitively addressed the legality of this practice. Given these legal hurdles, Runway’s collaboration with Lionsgate is a tactical decision to ensure that its training data is fully sanctioned and legally sound.

This collaboration may serve as a template for other AI enterprises seeking to sidestep legal complications by forming alliances with content proprietors. By leveraging Lionsgate’s proprietary material, Runway can develop its AI models without facing the risk of copyright infringements, potentially establishing a new norm for how AI companies handle data acquisition.

### The Industry’s Divided Opinions

Although the alliance between Lionsgate and Runway has the potential to transform content production, it has also initiated considerable discourse within the entertainment sector. On one side, studios may perceive AI as a means to streamline production, cut costs, and foster creativity. On the flip side, many industry professionals—especially unions, actors, and artists—express apprehension regarding the impact of AI on job stability and creative authenticity.

Critics warn that AI may supplant human artists and filmmakers, prompting ethical concerns about the industry’s future. For instance, filmmaker and outspoken AI critic Joe Russo shared his discontent on social media, stating, “I don’t think I’ve ever encountered a more repugnant phrase than: ‘to develop cutting-edge, capital-efficient content creation opportunities.'” Similarly, *Hunger Games* concept artist Reid Southen articulated his discontent, asserting, “I wonder how the directors and actors of their films feel about their work being used to develop the AI model. As an artist on *The Hunger Games*? I’m furious. This is the initial step toward attempting to replace artists and filmmakers.”

These worries are not baseless. AI’s capability to generate images, videos, and complete scenes from text prompts could potentially lessen the need for human labor in specific filmmaking facets. This has raised concerns that AI might displace jobs in areas such as visual effects, animation, and scriptwriting. Furthermore, there are apprehensions about the unauthorized use of actors’ likenesses, with some fearing that AI could create digital replicas of performers without their consent.

### The Future of AI in Hollywood

In spite of these apprehensions, the advancement of AI video synthesis technology is surging forward, and its integration into Hollywood appears

Congress Unveils Two Legislative Proposals Targeting the Reinstatement of Patent Safeguards for Software and Genetic Sequences

### Senate Judiciary Committee Weighs Bills to Repeal Supreme Court Decisions on Software and Gene Patents

The Senate Judiciary Committee is poised to examine two crucial bills that could transform the patent law framework in the United States. Should these bills be enacted, they would effectively invalidate important Supreme Court decisions that have curtailed patents on extensive software processes and human genes. The legislative initiatives have ignited a vigorous discussion, with advocates of open-source, biotech companies, and pharmaceutical firms all sharing their perspectives on the potential repercussions.

#### The Patent Eligibility Restoration Act (PERA)

One of the proposals being evaluated is the **Patent Eligibility Restoration Act** (PERA, S. 2140), introduced by Senators Thom Tillis (R-NC) and Chris Coons (D-Del.). This legislation seeks to revise the U.S. Code to abolish all judicial exceptions to patent eligibility, which would entail reversing several pivotal Supreme Court verdicts.

A significant case that would be impacted by PERA is the **2014 Supreme Court decision** in *Alice Corp. v. CLS Bank International*. In this ruling, the Court determined that simply executing an existing procedure on a computer does not render it a new, patentable invention. Justice Clarence Thomas, writing for the majority, asserted that abstract concepts applied to generic computers are not patentable. This ruling has been fundamental in constraining excessively broad software patents, particularly those that merely automate pre-existing business methods.

The *Alice* decision built on the prior case of *Bilski v. Kappos* (2010), where the Court denied a patent application for a method of hedging risks in commodity trading. Both decisions have played key roles in preventing what many view as unwarranted software patents, which often hinder innovation and result in expensive legal disputes.

#### Open Source Advocates Rally

The proposal of PERA has alarmed advocates of open-source and internet freedom. Organizations such as the **Linux Foundation** and the **Cloud Native Computing Foundation (CNCF)** have voiced their worries that the bill could lead to a revival of broad, ambiguous software patents, posing a threat to the open-source community.

In reaction, the Linux Foundation and CNCF have strengthened their collaboration with **Unified Patents**, a firm committed to protecting open-source software from patent trolls—entities that hold patents primarily to extract settlements instead of creating products. According to Unified Patents, nearly 98% of claims made by these “non-practicing entities” (NPEs) are settled outside of court, often due to the prohibitive costs of litigation. Nonetheless, when these claims are contested at the U.S. Patent and Trademark Appeals Board, NPEs face a defeat rate of 67%.

The open-source sector is concerned that PERA could incentivize patent trolls, resulting in a rise in baseless lawsuits that could hinder innovation and impose legal costs on small developers.

#### The PREVAIL Act: Restricting Patent Challenges

Another bill under consideration is the **PREVAIL Act** (S. 2220), which could complicate the process for disputing questionable patents. This legislation would set a 14,000-word cap on petitions challenging patents, which may obstruct efforts to dismantle intricate and overly broad patents. Furthermore, the PREVAIL Act would eliminate “clearance patents,” which businesses utilize to verify that their products do not breach existing patents before launching in the market.

Critics contend that these alterations would make it increasingly difficult to challenge poor patents, thereby bolstering patent trolls and large corporations with extensive patent portfolios. The **Electronic Frontier Foundation (EFF)** has been particularly outspoken against both PERA and PREVAIL, labeling them as “a significant boon to patent trolls, a select few tech firms that aggressively license patents, and patent attorneys.”

#### Genetic Patents: A Regressive Move?

Beyond software patents, PERA would also affect the patentability of human genes. In **2013**, the Supreme Court ruled in *Association for Molecular Pathology v. Myriad Genetics* that naturally occurring DNA sequences cannot be patented, even if they have been isolated from the human genome. This ruling was a major success for supporters of affordable healthcare, as it invalidated patents held by Myriad Genetics on the BRCA1 and BRCA2 genes, which are associated with breast and ovarian cancer.

If PERA passes, companies could again patent isolated human genes, potentially establishing monopolies on genetic testing for specific diseases. This has raised alarms among healthcare advocates, who fear such patents could escalate the cost of genetic testing and restrict access to vital diagnostics.

#### The Discussion on Innovation and Patent Law

Supporters of PERA and PREVAIL, including pharmaceutical trade groups like **PhRMA**, argue that the measures are essential to safeguard innovation, particularly in emerging sectors like biotechnology. Senator Thom Tillis informed **Axios** that the bills would offer relief to industries dependent on patents to shield their investments in