Welcome to this week’s Threat Source newsletter.
This week, we explore how advances in agent AI are rapidly transforming the cybercrime business.
Agenttic AI programming gives autonomy to AI agents, allowing them to interact with external systems to gather information, make decisions with the help of generative AI systems, and bring about changes in the external environment. Activities are executed via various APIs in the context of a defined workflow according to instructions provided to the agent.
The benefit for human operators is that these systems can efficiently perform routine tasks that would normally require access to multiple systems. Essentially, the AI agent acts as a trusted assistant, allowing human operators to focus on other things while completing tasks with minimal supervision.
Just as this approach brings benefits to the legitimate economy, it also brings similar efficiencies to the cybercrime economy. Recently, publications The discovery of the first cyber campaign orchestrated by AI should give us pause. This heralds a new era for cybersecurity teams.
We are entering an era where we can expect a lot of experimentation and innovation with AI in both the legal and cybercrime economies. AI acts as a force enabler, making tasks easier and faster to perform. Similarly, AI lowers barriers to entry and allows less skilled actors to perform tasks that they lack the skill to perform. AI does not bring new capabilities, but it can make existing functions easier to perform. However, AI systems still require skilled guidance and supervision.
AI is not infallible and tends to get things wrong and make up nonsense. When things go off the rails, humans must intervene to resolve the situation. This is not always easy and can be difficult for less skilled attackers.
Don’t be discouraged. You can also use these developments to your advantage. Defense teams can create their own agent systems to find and fix weaknesses in a system before a malicious attacker can identify them. You can deploy honeypot systems designed to be discovered by malicious AI systems, engage with them, and tie up their resources.
The threat landscape is never static. While AI makes some tasks more accessible to threat actors, it is a double-edged sword and also presents opportunities for defenders.
one big thing
Cisco Talos includes: Introduced new features For Snort3 users within Cisco Secure Firewall. The new Severity rule group allows you to organize detection rules by CVSS-based vulnerability severity: low, medium, high, or critical. This allows teams to prioritize and manage rules according to risk and urgency. You can also select rules based on the age of the vulnerability, such as the past 2, 5, or 10 years.
Why do I care?
This update gives you more flexibility and control. This makes it easy to maintain consistent and targeted detection coverage whether you’re running a large distributed network or a smaller environment with customized security priorities.
So what do you do now?
Review your current Snort3 rule configuration on Cisco Secure Firewall and consider adopting the new severity and time-based grouping features. By tailoring rule sets to your organization’s unique risk tolerance and patching cycles, you can optimize detection coverage, streamline management, and better protect your environment.
This week’s top security headlines
deadly railway braking system open to falsification
Researchers have discovered a way to spoof the signal that tells train conductors to brake, opening the door to all sorts of dangerous attack scenarios. (dark reading)
echogram Defect bypasses key LLM guardrails
The flaw, discovered in early 2025 and named EchoGram, allows a specially chosen simple word or code sequence to completely fool the automatic defenses, or guardrails, that keep an AI safe. (hack lead)
Over 67,000 fakes npm Packages flood the registry with worm-like spam attacks
The nematode’s reproductive mechanism and unique naming scheme that relies on Indonesian names and food terminology for newly created packaging has earned it the nickname IndonesiaFoods Worm. The fake package pretends to be a Next.js project. (hacker news)
120,000 resumes leaked in Cornerstone Staffing ransomware attack, claims giraffe gang
The notorious Qilin gang posted an industry-leading recruitment agency on the Dark Creek blog last Thursday. The group claims to have leaked 300GB of sensitive information from Cornerstone. (cyber news)
surveillance technology provider protein Hacked, data stolen, website defaced
It’s unclear exactly when and how Prorei was hacked, but a copy of the company’s website preserved in the Internet Archive’s Wayback Machine shows it was defaced on November 8th and restored soon after. (tech crunch)
Not enough Talos?
TTP: How Talos built an AI model into one of the internet’s most abused layers
Hazel talks with Talos researcher David Rodriguez about how attackers use DNS tunneling to sneak data out of networks, why it’s so difficult to discover in real time, and how Talos built an AI model to detect attackers without destroying anything important (like the internet).
Humanity of Talos: About epic reading, lifelong learning, and empathy.
In this episode, Bill Larget talks about what drew him to Talos, how his love of reading shaped his cyber security ethos, and the key insights he shares with the next generation of cyber security professionals.
Unleashing the Kraken ransomware group
In August 2025, Cisco Talos observed a big game hunt and double extortion attack carried out by Kraken, a Russian-speaking group that emerged from the remnants of the HelloKitty ransomware cartel.
Upcoming events where you can find Talos
- Deepsec IDSC (November 18th-21st) Vienna, Austria
- Aval (December 3rd to 5th) Kuala Lumpur, Malaysia
Most prevalent malware files from Talos telemetry in the past week
SHA256: 9f1f11a708d393e0a4109ae189bc64f1f3e312653dcf317a2bd406f18ffcc507
MD5: 2915b3f8b703eb744fc54c81f4a9c67f
Talos person in charge: https://talosintelligence.com/talos_file_reputation?s=9f1f11a708d393e0a4109ae189bc64f1f3e312653dcf317a2bd406f18ffcc507
Example file name: e74d9994a37b2b4c693a76a580c3e8fe_1_Exe.exe
Detection name: Win.Worm.Coinminer::1201
SHA256: 90b1456cdbe6bc2779ea0b4736ed9a998a71ae37390331b6ba87e389a49d3d59
MD5: c2efb2dcacba6d3ccc175b6ce1b7ed0a
Talos person in charge: https://talosintelligence.com/talos_file_reputation?s=90b1456cdbe6bc2779ea0b4736ed9a998a71ae37390331b6ba87e389a49d3d59
Example file name: ck8yh2og.dll
Detection name: Auto.90B145.282358.in02
SHA256: 96fa6a7714670823c83099ea01d24d6d3ae8fef027f01a4ddac14f123b1c9974
MD5: aac3165ece2959f39ff98334618d10d9
Talos person in charge: https://talosintelligence.com/talos_file_reputation?s=96fa6a7714670823c83099ea01d24d6d3ae8fef027f01a4ddac14f123b1c9974
Example file name: 96fa6a7714670823c83099ea01d24d6d3ae8fef027f01a4ddac14f123b1c9974.exe
Detection name: W32.Injector:Gen.21ie.1201
SHA256: c0ad494457dcd9e964378760fb6aca86a23622045bca851d8f3ab49ec33978fe
MD5: bf9672ec85283fdf002d83662f0b08b7
Talos person in charge: https://talosintelligence.com/talos_file_reputation?s=c0ad494457dcd9e964378760fb6aca86a23622045bca851d8f3ab49ec33978fe
Example file name: f_0008d7.html
Detected name: W32.C0AD494457-95.SBX.TG
SHA256: a31f222fc283227f5e7988d1ad9c0aecd66d58bb7b4d8518ae23e110308dbf91
MD5: 7bdbd180c081fa63ca94f9c22c457376
Talos person in charge: https://talosintelligence.com/talos_file_reputation?s=a31f222fc283227f5e7988d1ad9c0aecd66d58bb7b4d8518ae23e110308dbf91
Example file name: e74d9994a37b2b4c693a76a580c3e8fe_3_Exe.exe
Detection name: Win.Dropper.Miner::95.sbx.tg
