Category: Cybersecurity

Security doesn’t fail at the point of breach. It fails at the point of impact.

That line set the tone for this year’s Picus Breach and Simulation (BAS) Summit, where researchers, practitioners, and CISOs all echoed the same theme: cyber defense is no longer about prediction. It’s about proof.

When a new exploit drops, scanners scour the internet in minutes. Once attackers gain a foothold, lateral movement often follows just as fast. If your controls haven’t been tested against the exact techniques in play, you’re not defending, you’re hoping things don’t go seriously pear-shaped.

That’s why pressure builds long before an incident report is written. The same hour an exploit hits Twitter, a boardroom wants answers. As one speaker put it, “You can’t tell the board, ‘I’ll have an answer next week.’ We have hours, not days.”

BAS has outgrown its compliance roots and become the daily voltage test of cybersecurity, the current you run through your stack to see what actually holds.

This article isn’t a pitch or a walkthrough. It’s a recap of what came up on stage, in essence, how BAS has evolved from an annual checkbox activity to a simple and effective everyday way of proving that your defenses are actually working.

Security isn’t about design, it’s about reaction

For decades, security was treated like architecture: design, build, inspect, certify. A checklist approach built on plans and paperwork.

Attackers never agreed to that plan, however. They treat defense like physics, applying continuous pressure until something bends or breaks. They don’t care what the blueprint says; they care where the structure fails.

Pentests still matter, but they’re snapshots in motion.

BAS changed that equation. It doesn’t certify a design; it stress-tests the reaction. It runs safe, controlled adversarial behaviors in live environments to prove whether defenses actually respond as they should or not.

As Chris Dale, Principal Instructor at SANS, explains: The difference is mechanical: BAS measures reaction, not potential. It doesn’t ask, “Where are the vulnerabilities?” but “What happens when we hit them?”

Because ultimately, you don’t lose when a breach happens, you lose when the impact of that breach lands.

sıla-blog-video-1_1920x1080.mp4

Then assume a breach and work backward from the outcome you fear the most.

Take Akira, for instance, a ransomware chain that deletes backups, abuses PowerShell, and spreads through shared drives. Replay that behavior safely inside your environment, and you’ll learn, not guess, whether your defenses can break it midstream.

Two principles separated mature programs from the rest:

• Outcome first: start from impact, not inventory.
• Purple by default: BAS isn’t red-versus-blue theater; it’s how intel, engineering, and operations converge — simulate → observe → tune → re-simulate.

As John Sapp, CISO at Texas Mutual Insurance noted, “teams that make validation a weekly rhythm start seeing proof where they used to see assumptions.”

The real work of AI is curation, not creation

AI was everywhere this year, but the most valuable insight wasn’t about power, it was about restraint. Speed matters, but provenance matters more. Nobody wants an LLM model improvising payloads or making assumptions about attack behavior.

For now, at least, the most useful kind of AI isn’t the one that creates, it’s the one that organizes, taking messy, unstructured threat intelligence and turning it into something defenders can actually use.

sıla-blog-video-2_1920x1080.mp4

AI now acts less like a single model and more like a relay of specialists, each with a specific job and a checkpoint in between:

• Planner — defines what needs to be collected.
• Researcher — verifies and enriches threat data.
• Builder — structures the information into a safe emulation plan.
• Validator — checks fidelity before anything runs.

Each agent reviews the last, keeping accuracy high and risk low.

One example summed it up perfectly:

“Give me the link to the Fin8 campaign, and I’ll show you the MITRE techniques it maps to in hours, not days.”

That’s no longer aspirational, it’s operational. What once took a week of manual cross-referencing, scripting, and validation now fits inside a single workday.

Headline → Emulation plan → Safe run. Not flashy, just faster. Again, hours, not days.

Proof from the field shows that BAS works

One of the most anticipated sessions of the event was a live showcase of BAS in real environments. It wasn’t theory, it was operational proof.

A healthcare team ran ransomware chains aligned with sector threat intel, measuring time-to-detect and time-to-respond, feeding missed detections back into SIEM and EDR rules until the chain broke early.

An insurance provider demonstrated weekend BAS pilots to verify whether endpoint quarantines actually triggered. Those runs exposed silent misconfigurations long before attackers could.

The takeaway was clear:

BAS is already part of daily security operations, not a lab experiment. When leadership asks, “Are we protected against this?” the answer now comes from evidence, not opinion.

Validation turns “patch everything” into “patch what matters”

One of the summit’s sharpest moments came when the familiar board question surfaced: “Do we need to patch everything?”

The answer was unapologetically clear, no.

sıla-blog-video-3_1920x1080.mp4

BAS-driven validation proved that patching everything isn’t just unrealistic; it’s unnecessary.

What matters is knowing which vulnerabilities are actually exploitable in your environment. By combining vulnerability data with live control performance, security teams can see where real risk concentrates, not where a scoring system says it should.

“You shouldn’t patch everything,” Volkan Ertürk, Picus Co-Founder & CTO said. “Leverage control validation to get a prioritized list of exposures and focus on what’s truly exploitable for you.”

A CVSS 9.8 shielded by validated prevention and detection may carry little danger, while a medium-severity flaw on an exposed system can open a live attack path.

That shift, from patching on assumption to patching on evidence, was one of the event’s defining moments. BAS doesn’t tell you what’s wrong everywhere; it tells you what can hurt you here, turning Continuous Threat Exposure Management (CTEM) from theory into strategy.

You don’t need a moonshot to start

Another key takeaway from Picus security architecture leaders Gürsel Arıcı and Autumn Stambaugh’s session was that BAS doesn’t require a grand rollout; it simply needs to get started.

Teams began without fuss or fanfare, proving value in weeks, not quarters.

• Most picked one or two scopes, finance endpoints, or a production cluster, and mapped the controls protecting them.
• Then they chose a realistic outcome, like data encryption, and built the smallest TTP chain that could make it happen.
• Run it safely, see where prevention or detection fails, fix what matters, and run it again.

In practice, that loop accelerated fast.

By week three, AI-assisted workflows were already refreshing threat intel and regenerating safe actions. By week four, validated control data and vulnerability findings merged into exposure scorecards that executives could read at a glance.

The moment a team watched a simulated kill chain stop mid-run because of a rule shipped the day before, everything clicked, BAS stopped being a project and became part of their daily security practice.

BAS works as the verb inside CTEM

Gartner’s Continuous Threat Exposure Management (CTEM) model: “Assess, validate, mobilize” only works when validation is continuous, contextual, and tied to action.

This is where BAS lives now.

It’s not a standalone tool; it’s the engine that keeps CTEM honest, feeding exposure scores, guiding control engineering, and sustaining agility as both your tech stack and the threat surface shift.

The best teams run validation like a heartbeat. Every change, every patch, every new CVE triggers another pulse. That’s what continuous validation actually means.

The future lies in proof

Security used to run on belief. BAS replaces belief with proof, running electrical current through your defenses to see where the circuit fails.

AI brings speed. Automation brings scale. Validation brings truth. BAS isn’t how you talk about security anymore. It’s how you prove it.

Be among the first to experience AI-powered threat intelligence. Get your early access now!

Note: This article was expertly written and contributed by Sila Ozeren Hacioglu, Security Research Engineer at Picus Security.

Oct 30, 2025Ravie LakshmananDevSecOps / Software Security

Cybersecurity researchers have uncovered yet another active software supply chain attack campaign targeting the npm registry with over 100 malicious packages that can steal authentication tokens, CI/CD secrets, and GitHub credentials from developers’ machines.

The campaign has been codenamed PhantomRaven by Koi Security. The activity is assessed to have begun in August 2025, when the first packages were uploaded to the repository. It has since ballooned to a total of 126 npm libraries, attracting more than 86,000 installs.

Some of the packages have also been flagged by the DevSecOps company DCODX –

• op-cli-installer (486 Downloads)
• unused-imports (1,350 Downloads)
• badgekit-api-client (483 Downloads)
• polyfill-corejs3 (475 Downloads)
• eslint-comments (936 Downloads)

What makes the attack stand out is the attacker’s pattern of hiding the malicious code in dependencies by pointing to a custom HTTP URL, causing npm to fetch them from an untrusted website (in this case, “packages.storeartifact[.]com”) as opposed to npmjs[.]com each time a package is installed.

“And npmjs[.]com doesn’t follow those URLs,” security researcher Oren Yomtov laid out in a report shared with The Hacker News. “Security scanners don’t fetch them. Dependency analysis tools ignore them. To every automated security system, these packages show ‘0 Dependencies.’”

More worryingly, the fact that the URL is attacker-controlled means that it can be abused by the bad actor to tailor their payloads and serve any kind of malware, and make it more stealthy by initially serving completely harmless code before pushing a malicious version of the dependency after the package gains broader adoption.

The attack chain kicks off as soon as a developer installs one of the “benign” packages, which, in turn, leads to the retrieval of the remote dynamic dependency (RDD) from the external server. The malicious package comes with a pre-install hook that triggers the execution of the main payload.

The malware is designed to scan the developer environment for email addresses, gather information about the CI/CD environment, collect a system fingerprint, including the public IP address, and exfiltrate the results to a remote server.

Koi Security said the choice of the package names is not random, and that the threat actor has resorted to capitalizing on a phenomenon called slopsquatting – where large language models (LLMs) hallucinate non-existent yet plausible-sounding package names – in order to register those packages.

“PhantomRaven demonstrates how sophisticated attackers are getting [better] at exploiting blind spots in traditional security tooling,” Yomtov said. “Remote Dynamic Dependencies aren’t visible to static analysis. AI hallucinations create plausible-sounding package names that developers trust. And lifecycle scripts execute automatically, without any user interaction.”

The development once again illustrates how threat actors are finding novel ways to hide malicious code in open-source ecosystems and fly under the radar.

“The npm ecosystem allows easy publishing and low friction for packages,” DCODX said. “Lifecycle scripts (preinstall, install, postinstall) execute arbitrary code at install time, often without developer awareness.”

Oct 30, 2025Ravie LakshmananCybersecurity / Hacking News

The comfort zone in cybersecurity is gone. Attackers are scaling down, focusing tighter, and squeezing more value from fewer, high-impact targets. At the same time, defenders face growing blind spots — from spoofed messages to large-scale social engineering.

This week’s findings show how that shrinking margin of safety is redrawing the threat landscape. Here’s what’s making headlines.

Cyber threats are evolving faster than most defenses can adapt, and the line between criminal enterprise and nation-state tactics keeps blurring. Staying ahead now means staying aware — of every small shift in tools, tradecraft, and targeting. Until next ThreatsDay, stay sharp and stay curious.

Oct 29, 2025Ravie LakshmananMachine Learning / AI Safety

Cybersecurity researchers have flagged a new security issue in agentic web browsers like OpenAI ChatGPT Atlas that exposes underlying artificial intelligence (AI) models to context poisoning attacks.

In the attack devised by AI security company SPLX, a bad actor can set up websites that serve different content to browsers and AI crawlers run by ChatGPT and Perplexity. The technique has been codenamed AI-targeted cloaking.

The approach is a variation of search engine cloaking, which refers to the practice of presenting one version of a web page to users and a different version to search engine crawlers with the end goal of manipulating search rankings.

The only difference in this case is that attackers optimize for AI crawlers from various providers by means of a trivial user agent check that leads to content delivery manipulation.

“Because these systems rely on direct retrieval, whatever content is served to them becomes ground truth in AI Overviews, summaries, or autonomous reasoning,” security researchers Ivan Vlahov and Bastien Eymery said. “That means a single conditional rule, ‘if user agent = ChatGPT, serve this page instead,’ can shape what millions of users see as authoritative output.”

SPLX said AI-targeted cloaking, while deceptively simple, can also be turned into a powerful misinformation weapon, undermining trust in AI tools. By instructing AI crawlers to load something else instead of the actual content, it can also introduce bias and influence the outcome of systems leaning on such signals.

“AI crawlers can be deceived just as easily as early search engines, but with far greater downstream impact,” the company said. “As SEO [search engine optimization] increasingly incorporates AIO [artificial intelligence optimization], it manipulates reality.”

The disclosure comes as an analysis of browser agents against 20 of the most common abuse scenarios, ranging from multi-accounting to card testing and support impersonation, discovered that the products attempted nearly every malicious request without the need for any jailbreaking, the hCaptcha Threat Analysis Group (hTAG) said.

Furthermore, the study found that in scenarios where an action was “blocked,” it mostly came down due to the tool missing a technical capability rather than due to safeguards built into them. ChatGPT Atlas, hTAG noted, has been found to carry out risky tasks when they are framed as part of debugging exercises.

Claude Computer Use and Gemini Computer Use, on the other hand, have been identified as capable of executing dangerous account operations like password resets without any constraints, with the latter also demonstrating aggressive behavior when it comes to brute-forcing coupons on e-commerce sites.

hTAG also tested the safety measures of Manus AI, uncovering that it executes account takeovers and session hijacking without any issue, while Perplexity Comet runs unprompted SQL injection to exfiltrate hidden data.

“Agents often went above and beyond, attempting SQL injection without a user request, injecting JavaScript on-page to attempt to circumvent paywalls, and more,” it said. “The near-total lack of safeguards we observed makes it very likely that these same agents will also be rapidly used by attackers against any legitimate users who happen to download them.”

Oct 29, 2025Ravie LakshmananVulnerability / Internet of Things

Cybersecurity researchers are calling attention to a spike in automated attacks targeting PHP servers, IoT devices, and cloud gateways by various botnets such as Mirai, Gafgyt, and Mozi.

“These automated campaigns exploit known CVE vulnerabilities and cloud misconfigurations to gain control over exposed systems and expand botnet networks,” the Qualys Threat Research Unit (TRU) said in a report shared with The Hacker News.

The cybersecurity company said PHP servers have emerged as the most prominent targets of these attacks owing to the widespread use of content management systems like WordPress and Craft CMS. This, in turn, creates a large attack surface as many PHP deployments can suffer from misconfigurations, outdated plugins and themes, and insecure file storage.

Some of the prominent weaknesses in PHP frameworks that have been exploited by threat actors are listed below –

• CVE-2017-9841 – A Remote code execution vulnerability in PHPUnit
• CVE-2021-3129 – A Remote code execution vulnerability in Laravel
• CVE-2022-47945 – A Remote code execution vulnerability in ThinkPHP Framework

Qualys said it has also observed exploitation efforts that involve the use of “/?XDEBUG_SESSION_START=phpstorm” query string in HTTP GET requests to initiate an Xdebug debugging session with an integrated development environment (IDE) like PhpStorm.

“If Xdebug is unintentionally left active in production environments, attackers may use these sessions to gain insight into application behavior or extract sensitive data,” the company said.

Alternatively, threat actors are continuing to look for credentials, API keys, and access tokens in internet-exposed servers to take control of susceptible systems, as well as leverage known security flaws in IoT devices to co-opt them into a botnet. These include –

• CVE-2022-22947 – A Remote code execution vulnerability in Spring Cloud Gateway
• CVE-2024-3721 – A Command injection vulnerability in TBK DVR-4104 and DVR-4216
• A Misconfiguration in MVPower TV-7104HE DVR that allows unauthenticated users to execute arbitrary system commands via an HTTP GET request

The scanning activity, Qualys added, often originates from cloud infrastructures like Amazon Web Services (AWS), Google Cloud, Microsoft Azure, Digital Ocean, and Akamai Cloud, illustrating how threat actors are abusing legitimate services to their advantage while obscuring their true origins.

“Today’s threat actors don’t need to be highly sophisticated to be effective,” it noted. “With widely available exploit kits, botnet frameworks, and scanning tools, even entry-level attackers can cause significant damage.”

To safeguard against the threat, it’s advised that users keep their devices up-to-date, remove development and debug tools in production environments, secure secrets using AWS Secrets Manager or HashiCorp Vault, and restrict public access to cloud infrastructure.

“While botnets have previously been associated with large-scale DDoS attacks and occasional crypto mining scams, in the age of identity security threats, we see them taking on a new role in the threat ecosystem,” James Maude, field CTO at BeyondTrust, said.

“Having access to a vast network of routers and their IP addresses can allow threat actors to perform credential stuffing and password spray attacks a huge scale. Botnets can also evade geolocation controls by stealing a user’s credentials or hijacking a browser session and then using a botnet node close to the victim’s actual location and maybe even using the same ISP as the victim to evade unusual login detections or access policies.”

The disclosure comes as NETSCOUT classified the DDoS-for-hire botnet known as AISURU as a new class of malware dubbed TurboMirai that can launch DDoS attacks that exceed 20 terabits per second (Tbps). The botnet primarily comprises consumer-grade broadband access routers, online CCTV and DVR systems, and other customer premise equipment (CPE).

“These botnets incorporate additional dedicated DDoS attack capabilities and multi-use functions, enabling both DDoS attacks and other illicit activities such as credential stuffing, artificial intelligence (AI)-driven web scraping, spamming, and phishing,” the company said.

“AISURU includes an onboard residential proxy service used to reflect HTTPS application-layer DDoS attacks generated by external attack harnesses.”

Turning compromised devices into a residential proxy allows paying customers to route their traffic through one of the nodes in the botnet, offering anonymity and the ability to blend in with regular network activity. According to independent security journalist Brian Krebs, all of the major proxy services have grown exponentially over the past six months, citing data from spur.us.

Organizations in Ukraine have been targeted by threat actors of Russian origin with an aim to siphon sensitive data and maintain persistent access to compromised networks.

The activity, according to a new report from the Symantec and Carbon Black Threat Hunter Team, targeted a large business services organization for two months and a local government entity in the country for a week.

The attacks mainly leveraged living-off-the-land (LotL) tactics and dual-use tools, coupled with minimal malware, to reduce digital footprints and stay undetected for extended periods of time.

“The attackers gained access to the business services organization by deploying web shells on public-facing servers, most likely by exploiting one or more unpatched vulnerabilities,” the Broadcom-owned cybersecurity teams said in a report shared with The Hacker News.

One of the web shells used in the attack was Localolive, which was previously flagged by Microsoft as put to use by a sub-group of the Russia-linked Sandworm crew as part of a multi-year campaign codenamed BadPilot. LocalOlive is designed to facilitate the delivery of next-stage payloads like Chisel, plink, and rsockstun. It has been utilized since at least late 2021.

Early signs of malicious activity targeting the business services organization date back to June 27, 2025, with the attackers leveraging the foothold to drop a web shell and use it to conduct reconnaissance. The threat actors have also been found to run PowerShell commands to exclude the machine’s Downloads from Microsoft Defender Antivirus scans, as well as set up a scheduled task to perform a memory dump every 30 minutes.

Over the next couple of weeks, the attackers carried out a variety of actions, including –

• Save a copy of the registry hive to a file named 1.log
• Dropping more web shells
• Using the web shell to enumerate all files in the user directory
• Running a command to list all running processes beginning with “kee,” likely with the goal of targeting the KeePass password storage vault
• Listing all active user sessions on a second machine
• Running executables named “service.exe” and “cloud.exe” located in the Downloads folder
• Running reconnaissance commands on a third machine and performing a memory dump using the Microsoft Windows Resource Leak Diagnostic tool (RDRLeakDiag)
• Modifying the registry permits RDP connections to allow inbound RDP connections
• Running a PowerShell command to retrieve information about the Windows configuration on a fourth machine
• Running RDPclip to gain access to the clipboard in remote desktop connections
• Installing OpenSSH to facilitate remote access to the computer
• Running a PowerShell command to allow TCP traffic on port 22 for the OpenSSH server
• Creating a scheduled task to run an unknown PowerShell backdoor (link.ps1) every 30 minutes using a domain account
• Running an unknown Python script
• Deploying a legitimate MikroTik router management application (“winbox64.exe“) in the Downloads folder

Interestingly, the presence of “winbox64.exe” was also documented by CERT-UA in April 2024 in connection with a Sandworm campaign aimed at energy, water, and heating suppliers in Ukraine.

Symantec and Carbon Black said it could not find any evidence in the intrusions to connect it to Sandworm, but said it “did appear to be Russian in origin.” The cybersecurity company also revealed that the attacks were characterized by the deployment of several PowerShell backdoors and suspicious executables that are likely to be malware. However, none of these artifacts have been obtained for analysis.

“While the attackers used a limited amount of malware during the intrusion, much of the malicious activity that took place involved legitimate tools, either Living-off-the-Land or dual-use software introduced by the attackers,” Symantec and Carbon Black said.

“The attackers demonstrated an in-depth knowledge of Windows native tools and showed how a skilled attacker can advance an attack and steal sensitive information, such as credentials, while leaving a minimal footprint on the targeted network.”

The disclosure comes as Gen Threat Labs detailed Gamaredon‘s exploitation of a now-patched security flaw in WinRAR (CVE-2025-8088, CVSS score: 8.8) to strike Ukrainian government agencies.

“Attackers are abusing #CVE-2025-8088 (WinRAR path traversal) to deliver RAR archives that silently drop HTA malware into the Startup folder – no user interaction needed beyond opening the benign PDF inside,” the company said in a post on X. “These lures are crafted to trick victims into opening weaponized archives, continuing a pattern of aggressive targeting seen in previous campaigns.”

The findings also follow a report from Recorded Future, which found that the Russian cybercriminal ecosystem is being actively shaped by international law enforcement campaigns such as Operation Endgame, shifting the Russian government’s ties with e-crime groups from passive tolerance to active management.

Further analysis of leaked chats has uncovered that senior figures within these threat groups often maintain relationships with Russian intelligence services, providing data, performing tasking, or leveraging bribery and political connections for impunity. At the same time, cybercriminal crews are decentralizing operations to sidestep Western and domestic surveillance.

While it’s been long known that Russian cybercriminals could operate freely as long as they do not target businesses or entities operating in the region, Kremlin appears to be now taking a more nuanced approach where they recruit or co-opt talent when necessary, turn a blind eye when attacks align with their interests, and selectively enforce laws when the threat actors become “politically inconvenient or externally embarrassing.”

Viewed in that the “dark covenant” is a combination of several things: a commercial enterprise, tool of influence and information acquisition, and also a liability when it threatens domestic stability or because of Western pressure.

“The Russian cybercriminal underground is fracturing under the dual pressures of state control and internal mistrust, while proprietary forum monitoring and ransomware affiliate chatter show increasing paranoia among operators,” the company noted in its third instalment of the Dark Covenant report.

BeyondTrust’s annual cybersecurity predictions point to a year where old defenses will fail quietly, and new attack vectors will surge.

Introduction

The next major breach won’t be a phished password. It will be the result of a massive, unmanaged identity debt. This debt takes many forms: it’s the “ghost” identity from a 2015 breach lurking in your IAM, the privilege sprawl from thousands of new AI agents bloating your attack surface, or the automated account poisoning that exploits weak identity verification in financial systems. All of these vectors—physical, digital, new, and old—are converging on one single point of failure: identity.

Based on analysis from BeyondTrust’s cybersecurity experts, here are three critical identity-based threats that will define the coming year:

1. Agentic AI Emerges as the Ultimate Attack Vector

By 2026, agentic AI will be connected to nearly every technology we operate, effectively becoming the new middleware for most organizations. The problem is that this integration is driven by a speed-to-market push that leaves cybersecurity as an afterthought.

This rush is creating a massive new attack surface built on a classic vulnerability: the confused deputy problem.

A “deputy” is any program with legitimate privileges. The “confused deputy problem” occurs when a low-privilege entity—like a user, account, or another application—tricks that deputy into misusing its power to gain high privileges. The deputy, lacking the context to see the malicious intent, executes the command or shares results beyond its original design or intentions.

Now, apply this to AI. An agentic AI tool may be granted least privilege access to read a user’s email, access a CI/CD pipeline, or query a production database. If that AI, acting as a trusted deputy, is “confused” by a cleverly crafted prompt from another resource, it can be manipulated into exfiltrating sensitive data, deploying malicious code, or escalating higher privileges on the user’s behalf. The AI is executing tasks it has permission for, but on behalf of an attacker who does not, and can elevate privileges based on the attack vector.

Defender Tip:

This threat requires treating AI agents as potentially privileged machine identities. Security teams must enforce strict least privilege, ensuring AI tools only have the absolute minimum permissions necessary for specific tasks. This includes implementing context-aware access controls, command filtering, and real-time auditing to prevent these trusted agents from becoming malicious actors by proxy.

2. Account Poisoning: The Next Evolution of Financial Fraud

In the coming year, expect a significant rise in “account poisoning”, where threat actors find new ways to insert fraudulent billers and payees into consumer and business financial accounts at scale.

This “poison” is driven by automation that allows for the creation of payees and billers, the requesting of funds, and linking to other online payment processing sources. This attack vector is particularly dangerous because it exploits weaknesses in online financial systems, leverages poor secrets management to attack in bulk, and uses automation to obfuscate the transactions.

Defender Tip:

Security teams must move beyond flagging individual account takeovers and focus on high-velocity, automated changes to payee and biller information. The key is implementing tighter diligence and identity confidence checks for any automated process that requests to modify these financial fields.

3. Ghosts in Your IAM: Historic Identity Compromises Catch Up

Many organizations are finally modernizing their identity and access management (IAM) programs, adopting new tools, like graph-based analytics, to map their complex identity landscapes. In 2026, these efforts will uncover skeletons in the closet: “ghost” identities from long-past solutions and breaches that were never detected.

These “backdated breaches” will reveal rogue accounts—some years old—that remain in active use. Because these compromises are older than most security logs, it may be impossible for teams to determine the full extent of the original breach.

Defender Tip:

This prediction underscores the long-standing failure of basic joiner-mover-leaver (JML) processes. The immediate takeaway is to prioritize identity governance and use modern identity graphing tools to find and eliminate these dormant, high-risk accounts before they are rediscovered by attackers.

Other Trends on the Radar

The Death of the VPN

For years, the VPN was the workhorse of remote access, but in modern remote access, VPN is a critical vulnerability waiting to be exploited. Threat actors have mastered VPN exploitation techniques, using credential harvesting and compromised appliances for persistent access. Using traditional VPNs for privileged access presents a risk that organizations can no longer afford.

The Rise of AI Veganism

As a cultural counterforce, 2026 will witness the rise of “AI veganism”, where employees or customers abstain from using artificial intelligence on principle. This movement, driven by ethical concerns over data sourcing, algorithmic bias, and environmental costs, will challenge the assumption that AI adoption is inevitable. Companies will have to navigate this resistance by offering transparent governance, human-first alternatives, and clear opt-outs. However, when it comes to cybersecurity, opting out of AI-driven defenses may be less of an option and could even shift liability back to the user.

An Identity-First Security Posture is Non-Negotiable

The common thread through these 2026 predictions is identity. The new AI attack surface is an identity-privilege problem, account poisoning is an identity verification problem, while backdated breaches are an identity lifecycle problem. As the perimeter widens, organizations must adopt an identity-first security posture by applying principles of least privilege and zero trust to every human and non-human identity.

Want to get a deeper look at all of BeyondTrust’s 2026 cybersecurity predictions? Read the full report here.

Note: This article was written and contributed by Morey J. Haber, Chief Security Advisor; Christopher Hills, Chief Security Strategist; and James Maude, Field Chief Technology Officer at BeyondTrust.

Oct 29, 2025The Hacker NewsArtificial Intelligence / Compliance

Artificial Intelligence (AI) is rapidly transforming Governance, Risk, and Compliance (GRC). It’s no longer a future concept—it’s here, and it’s already reshaping how teams operate.

AI’s capabilities are profound: it’s speeding up audits, flagging critical risks faster, and drastically cutting down on time-consuming manual work. This leads to greater efficiency, higher accuracy, and a more proactive GRC function.

However, this powerful shift introduces significant new challenges. AI brings its own set of risks, including potential bias, dangerous blind spots, and regulatory gaps that are only beginning to be addressed by governing bodies. Staying ahead of this curve—not just struggling to keep up—requires clear, practical knowledge.

Don’t Just Stay Afloat—Master the Change

To help you navigate this complex landscape, we invite you to our free, high-impact webinar, “The Future of AI in GRC: Opportunities, Risks, and Practical Insights.” This session is designed to deliver clarity and direction for everyone, from those just starting out to teams actively scaling AI in their processes. It’s packed with practical advice—no fluff, no hype.

What You Will Learn

We will dive into the most critical aspects of AI in GRC, providing actionable takeaways you can implement immediately:

• Real-world examples of AI successfully improving compliance workflows.
• Early lessons and best practices from teams leveraging advanced agentic AI.
• The most common risks teams overlook—and concrete strategies to spot and mitigate them.
• A clear view of what’s next in AI for GRC and how to strategically prepare your team.

The speed of AI innovation is immense, and new regulations are struggling to catch up. The growing gap between technological capability and legal framework represents your immediate risk exposure.

This webinar cuts through the complexity by bringing together experts, actionable examples, and real talk. You don’t have to wait until you’re forced to react to a risk; be the leader who is prepared for it.

Ready to confidently move forward and make AI a real competitive advantage in your compliance strategy? Register here—it’s free and filling fast, last call to join live.

Oct 29, 2025Ravie LakshmananMalware / Threat Intelligence

Cybersecurity researchers have discovered a set of 10 malicious npm packages that are designed to deliver an information stealer targeting Windows, Linux, and macOS systems.

“The malware uses four layers of obfuscation to hide its payload, displays a fake CAPTCHA to appear legitimate, fingerprints victims by IP address, and downloads a 24MB PyInstaller-packaged information stealer that harvests credentials from system keyrings, browsers, and authentication services across Windows, Linux, and macOS,” Socket security researcher Kush Pandya said.

The npm packages were uploaded to the registry on July 4, 2025, and accumulated over 9,900 downloads collectively –

• deezcord.js
• dezcord.js
• dizcordjs
• etherdjs
• ethesjs
• ethetsjs
• nodemonjs
• react-router-dom.js
• typescriptjs
• zustand.js

The multi-stage credential theft operation manifested in the form of various typosquatted packages impersonating popular npm libraries such as TypeScript, discord.js, ethers.js, nodemon, react-router-dom, and zustand.

Once installed, the malware serves a fake CAPTCHA prompt and displays authentic-looking output that mimics legitimate package installations to give the impression that the setup process is proceeding along expected lines. However, in the background, the package captures the victim’s IP address, sends it to an external server (“195.133.79[.]43”), and then proceeds to drop the main malware.

In each package, the malicious functionality is automatically triggered upon installation by means of a postinstall hook, launching a script named “install.js” that detects the victim’s operating system and launches an obfuscated payload (“app.js”) in a new Command Prompt (Windows), GNOME Terminal or x-terminal-emulator (Linux), or Terminal (macOS) window.

“By spawning a new terminal window, the malware runs independently of the npm install process,” Pandya noted. “Developers who glance at their terminal during installation see a new window briefly appear, which the malware immediately clears to avoid suspicion.”

The JavaScript contained within “app.js” is hidden through four layers of obfuscation — such as XOR cipher with a dynamically generated key, URL-encoding of the payload string, and using hexadecimal and octal arithmetic to obscure program flow — that are designed to resist analysis.

The end goal of the attack is to fetch and execute a comprehensive information stealer (“data_extracter”) from the same server that’s equipped to thoroughly scan the developer’s machine for secrets, authentication tokens, credentials, and session cookies from web browsers, configuration files, and SSH keys.

The stealer binary also incorporates platform-specific implementations to extract credentials from the system keyring using the keyring npm library. The harvested information is compressed into a ZIP archive and exfiltrated to the server.

“System keyrings store credentials for critical services including email clients (Outlook, Thunderbird), cloud storage sync tools (Dropbox, Google Drive, OneDrive), VPN connections (Cisco AnyConnect, OpenVPN), password managers, SSH passphrases, database connection strings, and other applications that integrate with the OS credential store,” Socket said.

“By targeting the keyring directly, the malware bypasses application-level security and harvests stored credentials in their decrypted form. These credentials provide immediate access to corporate email, file storage, internal networks, and production databases.”

Oct 29, 2025Ravie LakshmananVulnerability / Malware

Threat actors are actively exploiting multiple security flaws impacting Dassault Systèmes DELMIA Apriso and XWiki, according to alerts issued by the U.S. Cybersecurity and Infrastructure Security Agency (CISA) and VulnCheck.

The vulnerabilities are listed below –

• CVE-2025-6204 (CVSS score: 8.0) – A code injection vulnerability in Dassault Systèmes DELMIA Apriso that could allow an attacker to execute arbitrary code.
• CVE-2025-6205 (CVSS score: 9.1) – A missing authorization vulnerability in Dassault Systèmes DELMIA Apriso that could allow an attacker to gain privileged access to the application.
• CVE-2025-24893 (CVSS score: 9.8) – An improper neutralization of input in a dynamic evaluation call (aka eval injection) in XWiki that could allow any guest user to perform arbitrary remote code execution through a request to the “/bin/get/Main/SolrSearch” endpoint.

Both CVE-2025-6204 and CVE-2025-6205 affect DELMIA Apriso versions from Release 2020 through Release 2025. They were addressed by Dassault Systèmes in early August.

Interestingly, the addition of the two shortcomings to the Known Exploited Vulnerabilities (KEV) catalog comes a little over a month after CISA flagged the exploitation of another critical flaw in the same product (CVE-2025-5086, CVSS score: 9.0), a week after the SANS Internet Storm Center detected in-the-wild attempts. It’s currently not known if these efforts are related.

VulnCheck, which detected exploitation attempts targeting CVE-2025-24893, said the vulnerability is being abused as part of a two-stage attack chain that delivers a cryptocurrency miner. According to CrowdSec and Cyble, the vulnerability is said to have been weaponized in real-world attacks as far back as March 2025.

“We observed multiple exploit attempts against our XWiki canaries coming from an attacker geolocated in Vietnam,” VulnCheck’s Jacob Baines said. “The exploitation proceeds in a two-pass workflow separated by at least 20 minutes: the first pass stages a downloader (writes a file to disk), and the second pass later executes it.”

The payload uses wget to retrieve a downloader (“x640”) from “193.32.208[.]24:8080” and write it to the “/tmp/11909” location. The downloader, in turn, runs shell commands to fetch two additional payloads from the same server –

• x521, which fetches the cryptocurrency miner located at “193.32.208[.]24:8080/rDuiQRKhs5/tcrond”
• x522, which kills competing miners such as XMRig and Kinsing, and launches the miner with a c3pool.org configuration

The attack traffic, per VulnCheck, originates from an IP address that geolocates to Vietnam (“123.25.249[.]88“) and has been flagged as malicious in AbuseIPDB for engaging in brute-force attempts as recently as October 26, 2025.

In light of active exploitation, users are advised to apply the necessary updates as soon as possible to safeguard against threats. Several Civilian Executive Branch (FCEB) agencies are required to remediate the DELMIA Apriso flaws by November 18, 2025.