Customer portal
Articles Tagged with

SOS Intelligence

"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 04 May 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-26322

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-26322

 


 

2. CVE-2026-24763

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-24763

 


 

3. CVE-2026-25157

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25157

 


 

4. CVE-2026-25253

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25253

 


 

5. CVE-2026-25475

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25475

 


 

6. CVE-2026-26329

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-26329

 


 

7. CVE-2026-27001

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-27001

 


 

8. CVE-2026-31431

In the Linux kernel, the following vulnerability has been resolved:

crypto: algif_aead – Revert to operating out-of-place

This mostly reverts commit 72548b093ee3 except for the copying of
the associated data.

There is no benefit in operating in-place in algif_aead since the
source and destination come from different mappings. Get rid of
all the complexity added for in-place operation and just copy the
AD directly.

https://nvd.nist.gov/vuln/detail/CVE-2026-31431

 


 

9. CVE-2026-27940

llama.cpp is an inference of several LLM models in C/C++. Prior to b8146, the gguf_init_from_file_impl() in gguf.cpp is vulnerable to an Integer overflow, leading to an undersized heap allocation. Using the subsequent fread() writes 528+ bytes of attacker-controlled data past the buffer boundary. This is a bypass of a similar bug in the same file – CVE-2025-53630, but the fix overlooked some areas. This vulnerability is fixed in b8146.

https://nvd.nist.gov/vuln/detail/CVE-2026-27940

 


 

10. CVE-2025-53630

llama.cpp is an inference of several LLM models in C/C++. Prior to b8146, the gguf_init_from_file_impl() in gguf.cpp is vulnerable to an Integer overflow, leading to an undersized heap allocation. Using the subsequent fread() writes 528+ bytes of attacker-controlled data past the buffer boundary. This is a bypass of a similar bug in the same file – CVE-2025-53630, but the fix overlooked some areas. This vulnerability is fixed in b8146.

https://nvd.nist.gov/vuln/detail/CVE-2025-53630

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten –

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1. 

https://nvd.nist.gov/vuln/detail/

 


 

2.

https://nvd.nist.gov/vuln/detail/

 


 

3.

https://nvd.nist.gov/vuln/detail/

 


 

4.

https://nvd.nist.gov/vuln/detail/

 


 

5.

https://nvd.nist.gov/vuln/detail/

 


 

6.

https://nvd.nist.gov/vuln/detail/

 


 

7.

https://nvd.nist.gov/vuln/detail/

 


 

8.

https://nvd.nist.gov/vuln/detail/

 


 

9.

https://nvd.nist.gov/vuln/detail/

 


 

10.

https://nvd.nist.gov/vuln/detail/

 


"SOS
Investigation, Opinion, The Dark Web

Dark Web Services: current Average Prices (2026 Update)

Dark Web Services: current Average Prices (2026 Update)

Introduction

Back in 2022, I published our first deep dive into dark web pricing. At the time, the landscape was already complex, but it was still possible to draw fairly clean lines between the categories of goods and services being traded. Four years on, those lines have blurred considerably.

The underground economy has matured. Prices have shifted, new product categories have emerged, and the operational sophistication of threat actors has increased significantly. Ransomware-as-a-Service is now an established business model. AI-generated phishing kits are being sold alongside traditional credential dumps. Crypto drainers have become a category in their own right. And stealer log subscriptions are now one of the fastest-growing products on the dark web.

13th May 3pm UK Time

Webinar: 2026 Dark Web Pricing Report

For this updated report, we conducted an exhaustive crawl using our own SOS Intelligence DARKSEARCH platform, scanning active dark web marketplaces, forums, and paste sites throughout Q1 2026. We supplemented this with direct marketplace access via Tor and cross-referenced our findings against published industry research from PrivacySharks, DeepStrike, Privacy Affairs, and Trustwave. This time around, we have also expanded our scope significantly, covering narcotics, firearms, counterfeit goods, cryptocurrency fraud tools, and forged documents alongside the traditional cyber-focused categories. The result is what I believe to be one of the most comprehensive snapshots of dark web pricing available today.

So whether you are a security researcher, an MSSP building threat briefings, or a CISO trying to quantify the risk exposure your organisation faces, this should give you a solid, data-backed foundation.

Methodology

Our approach follows the intelligence cycle: direction, collection, processing, analysis, and dissemination. The direction phase was straightforward: understand what is being sold on the dark web in 2026 and at what price points.

For collection, we used the SOS Intelligence API v2 to run targeted keyword searches across our indexed dark web corpus. This includes content from over 50 active marketplaces, forums, paste sites, and Telegram channels. We queried across 20+ distinct product categories, including stolen financial instruments, identity documents, hacking services, malware, access brokers, narcotics (cocaine, heroin, methamphetamine, cannabis, MDMA, and prescription drugs), firearms, counterfeit goods, cryptocurrency fraud tools, and forged documents. We also accessed active Tor marketplaces directly to verify listed prices against real product pages. Key marketplaces analysed include Tor Market, Tor Amazon, Abacus Market, TheBreakingBad, Gun and Shell Factory, and several standalone vendor storefronts.

During processing and analysis, we normalised prices to USD (where vendors listed in EUR, GBP, or cryptocurrency) and calculated averages across multiple vendors where possible. Where a product category had significant variance (for example, initial access pricing can range from $500 to $50,000+), we present the typical range rather than a misleading average.

One thing worth noting: prices on the dark web are not static. They fluctuate based on supply, demand, law enforcement activity, and even seasonal patterns. What we present here is a snapshot, accurate to Q1 2026, and should be treated as indicative rather than definitive.

Stolen Financial Instruments

Financial data remains the bread and butter of dark web commerce. Credit card data, bank account credentials, and payment platform logins continue to dominate marketplace listings. The availability is enormous, driven in large part by the explosion in stealer log infections and large-scale data breaches.

Credit card data with CVV (card-not-present fraud) remains cheap and abundant. A single card with CVV typically sells for $10 to $40, depending on the issuing bank, card type, and associated balance. Cards with higher balances or from premium issuers command a premium. Cloned physical cards with PIN are a different proposition entirely, typically ranging from $100 for a single card with a $2,500 to $3,500 balance, up to $600 for a batch of 10 cards with a combined balance of $33,000 to $35,000.

ProductPrice Range (USD)Source/Notes
Credit card with CVV$10 – $40Per card, card-not-present
Cloned card with PIN (single)$100 – $250$2,500 – $3,500 balance
Cloned cards (batch of 10)$450 – $600$30,000 – $35,000 combined
AMEX Prepaid (EUR 2,500)$105 – $510Price varies by vendor
Bank login (US)$35 – $500Depends on bank and balance
Bank login (UK/EU)$50 – $1,000+Premium for verified accounts
Bank transfer service ($10K)$500Vendor guarantees delivery
PayPal account (verified)$15 – $55Balance $2,500 – $25,000
Crypto exchange account (Kraken)$249Fully verified
Crypto exchange account (general)$90 – $250Coinbase, Binance, etc.

Compared to our 2022 findings, the average price of stolen credit card data has dropped slightly, reflecting oversupply. Bank account credentials, on the other hand, have held steady or increased, particularly for UK and EU accounts where strong customer authentication (SCA) requirements make compromised credentials more valuable to attackers who can bypass these controls.

Identity Documents and Fullz

Fullz, complete identity packages containing name, date of birth, SSN, address, and often more, remain a staple of dark web commerce. The pricing here has been remarkably consistent over the years, which suggests stable supply chains, likely fed by the steady stream of data breaches affecting organisations globally.

Bulk purchasing drives the per-unit cost down significantly. A batch of 1,000 Social Security Numbers was listed on Tor Market at $65. Business fullz (company identity packages with EIN numbers) go for around $95 for a set of 10, which is particularly concerning for organisations worried about business identity theft and fraudulent corporate filings.

ProductPrice Range (USD)Notes
Individual fullz (US)$20 – $100Per identity, includes SSN
SSN batch (1,000 records)$65Bulk purchase, specific regions
Business fullz with EIN (10 pack)$95Corporate identity packages
US passport scan$100Digital copy only
US physical passport (forged)$3,000 – $3,800High-quality forgery
UK driver’s licence (forged)$500Physical document
EU national ID (forged)$300 – $700Varies by country
Medical records$50 – $500+Depends on completeness
Selfie with ID (for KYC bypass)$50 – $100Growing demand

A notable trend in 2026 is the growing market for KYC bypass packages. These typically include a stolen identity paired with a matching selfie (often obtained from stealer logs that capture webcam images), sold specifically to bypass Know Your Customer verification on financial platforms. This is a direct response to tightened identity verification requirements, and it represents an uncomfortable escalation in the identity fraud ecosystem.

DDoS and Hacking Services

DDoS-for-hire remains one of the most accessible attack services on the dark web. Entry-level DDoS attacks can be purchased for as little as $10 per hour, making it trivially cheap for anyone with a grudge and a cryptocurrency wallet. Monthly subscription packages for sustained DDoS capability go up to $850, though most listings cluster around $200 to $500 per month.

Hacking services for hire are more variable in pricing, reflecting the range of complexity involved. Simple social media account compromises sit at the lower end, while corporate network penetration and database extraction commands significantly higher fees.

ServicePrice Range (USD)Notes
DDoS attack (per hour)$10 – $50Basic layer 4/7 attack
DDoS subscription (monthly)$200 – $850Sustained capability
Social media account hack$25 – $100Facebook, Instagram, etc.
Email account compromise$100 – $500Corporate email higher
Website hacking$200 – $3,000Depends on target complexity
Corporate network access$500 – $10,000+Overlaps with IAB market
Phone hacking/spyware install$300 – $1,500Remote installation
Doxing service$25 – $200Varies by depth of research

The DDoS market has become increasingly commoditised. In 2022 we reported an average DDoS service price of around $382. That number has come down, driven by competition between providers and the proliferation of botnet infrastructure. The real concern is not the price itself but how easy it has become to launch these attacks with minimal technical knowledge.

Malware, Exploit Kits, and Phishing

This is where the dark web economy has seen some of its most significant evolution since our last report. The malware-as-a-service model is now firmly established, with vendors offering everything from basic RATs (Remote Access Trojans) through to sophisticated banking trojans and zero-day exploits.

Phishing kits have become particularly interesting. AI-generated phishing templates are now being sold at a premium, with vendors marketing their kits as capable of bypassing modern email security filters. The quality of these templates has improved dramatically, making traditional security awareness training less effective than it was even two years ago.

ProductPrice Range (USD)Notes
RAT (Remote Access Trojan)$45 – $500Off-the-shelf, basic features
Banking trojan$500 – $1,800Targeted at specific banks
Ransomware-as-a-Service kit$500 – $5,000Includes builder and panel
Stealer log subscription$100 – $1,024/moRedline, Raccoon, Vidar
Phishing kit (standard)$50 – $300Includes templates and hosting
Phishing kit (AI-generated)$200 – $800Bypass modern filters
Zero-day exploit (general)$5,000 – $200,000+Price varies enormously
Exploit kit (browser)$100 – $2,000Pre-packaged exploitation
Botnet rental (1,000 bots)$50 – $200/dayFor spam or DDoS
Keylogger$25 – $150Basic to advanced features

The Ransomware-as-a-Service (RaaS) market deserves special attention. Our platform currently tracks over 100 active ransomware groups, many of which operate affiliate programmes where the actual ransomware deployment is carried out by affiliates who pay a percentage (typically 20-30%) of the ransom to the RaaS operator. The barrier to entry for launching a ransomware campaign has never been lower, and this is reflected in the sustained growth of ransomware incidents globally.

Crypto Drainers and Mixers

This is a category that barely existed in 2022 and is now a significant segment of the dark web economy. Crypto drainers are tools designed to empty cryptocurrency wallets, typically deployed via phishing sites that mimic legitimate Web3 platforms and trick users into connecting their wallets and signing malicious transactions.

Our DARKSEARCH data turned up active listings on DNA Forums and other threat actor communities, with prices ranging from $50 for basic tutorials through to $1,000 for fully operational Solana drainer toolkits. The Tron Trap drainer tool was listed at $300, while general-purpose drainer kits sat around $200 to $500.

ProductPrice Range (USD)Notes
Crypto drainer kit (general)$200 – $500Multi-chain support
Solana drainer$1,000Chain-specific tooling
Tron Trap drainer$300Listed on DNA Forums
Crypto drainer tutorial$50 – $100DIY approach
Crypto mixing/tumbling service1-3% of amountPer transaction fee
Crypto cashout service15-25% of amountConversion to fiat

The emergence of drainer-as-a-service mirrors the RaaS model. Operators provide the tooling and infrastructure, affiliates drive traffic to phishing sites, and profits are split. Some drainer operators take a 20-30% cut of every wallet drained. For context, wallet drainer attacks stole hundreds of millions of dollars in cryptocurrency during 2025 alone, making this one of the highest-growth criminal sectors.

Initial Access Brokers

Initial Access Brokers (IABs) continue to be a critical part of the threat landscape. These are threat actors who specialise in gaining access to corporate networks and then selling that access to other criminals, typically ransomware operators. The IAB market is essentially the supply chain for ransomware.

Pricing varies enormously based on the target organisation’s size, industry, and the type of access being sold. VPN credentials for a small company might go for $500, while domain admin access to a large enterprise can command $10,000 or more. Our 2022 report found an average of around $7,700 for initial network access. In 2026, the range has widened as the market has matured, but the median sits around $2,000 to $5,000 for mid-market targets.

Access TypePrice Range (USD)Notes
VPN credentials (SME)$200 – $1,000Single organisation
RDP access (dedicated server)$10 – $100Commodity pricing
Domain admin (enterprise)$5,000 – $50,000+High-value targets
Web shell access$50 – $500Depends on target
cPanel/hosting access$10 – $50Bulk available
Database access (customer data)$500 – $10,000Depends on record count
Cloud infrastructure access$1,000 – $20,000AWS, Azure, GCP

Cloud infrastructure access is the emerging high-value category here. As organisations continue their migration to cloud platforms, compromised cloud credentials have become increasingly sought after. A set of AWS root account credentials for an enterprise can be worth significantly more than traditional on-premise network access, reflecting the potential blast radius of a cloud compromise.

Stolen Accounts and Subscriptions

The market for compromised online accounts remains massive, covering everything from streaming services to social media to gaming platforms. These are largely driven by credential stuffing attacks leveraging the billions of username/password pairs available from historical breaches, combined with the output of stealer log infections.

Account TypePrice Range (USD)Notes
Netflix/Disney+/streaming$4 – $25Per account, often shared
Spotify Premium$3 – $10Bulk available
Facebook account$25 – $45Higher for aged accounts
Instagram account$25 – $45Followers affect price
LinkedIn Premium$30 – $50Professional accounts
Gaming accounts (Steam, Epic)$10 – $100Game library affects price
Food delivery (Uber Eats, etc.)$5 – $20With stored payment
Email accounts (bulk)$2 – $10Per account
VPN service accounts$5 – $15NordVPN, ExpressVPN, etc.

What strikes me about this category is how cheap everything is. A Netflix account for $4, a Facebook account for $25. The low prices reflect the sheer volume of compromised credentials available. For most consumers, the inconvenience of having an account compromised is minor. But for organisations, compromised employee accounts, particularly email and LinkedIn, can be the starting point for targeted social engineering campaigns.

Counterfeit Currency and Documents

Counterfeit physical currency continues to be traded, though the market has evolved. Our crawl of Robinhood Market found fake Euro banknotes listed from $300 for a small batch up to $1,200 for larger quantities. Western Union transfer services were listed at $200 for a $2,000 transfer, representing a 10% fee.

Bank cheque templates have also become a notable category, with templates available from as little as $5 for basic designs up to $600 for comprehensive kits that include matching security features and printing instructions.

ProductPrice Range (USD)Notes
Counterfeit EUR banknotes$300 – $1,200Various denominations
Counterfeit USD banknotes$350 – $1,500Quality varies significantly
Western Union transfer ($2,000)$20010% fee structure
MoneyGram transfer$150 – $300Similar fee structure
Bank cheque templates$5 – $600Including security features
Counterfeit branded goods (guides)$20 – $200Manufacturing instructions

In our 2022 report, counterfeit currency averaged around $396 per $1,000 face value. The current rates are broadly similar, suggesting this market has reached a stable equilibrium. The real shift is towards digital fraud, with physical counterfeiting becoming a smaller proportion of overall dark web commerce.

Proxy and Hosting Infrastructure

Bulletproof hosting and residential proxy services continue to be essential infrastructure for cybercriminal operations. These services provide the anonymous, abuse-tolerant hosting that enables everything from phishing campaigns to command and control servers.

ServicePrice Range (USD)Notes
Bulletproof hosting (monthly)$50 – $500Abuse-tolerant, offshore
Residential proxy (monthly)$200 – $645Pool of residential IPs
SOCKS5 proxy (per IP)$1 – $10Single use or short-lived
VPN service (criminal-oriented)$5 – $30/moNo-log guarantees
Dedicated server (offshore)$100 – $400/moFull admin access
Domain + hosting bundle$20 – $100For phishing campaigns

Residential proxy pricing has actually increased since 2022, when we reported an average of $645 per month. The current range starts lower but premium services now charge more, reflecting growing demand from threat actors who need residential IP addresses to bypass fraud detection systems and CAPTCHAs.

AI-Enabled Criminal Services

This is entirely new territory since our 2022 report. The commoditisation of large language models has created a new category of criminal tooling that simply did not exist four years ago. Dark web forums now host discussions and sales of jailbroken AI models, custom-trained chatbots for social engineering, and AI-powered tools for generating convincing phishing content at scale.

While we did not find as many standardised price points for AI services as for other categories (the market is still maturing), the trend is clear. AI is being integrated into existing criminal workflows, particularly around social engineering, phishing content generation, and code development for malware. Some vendors are marketing “FraudGPT” and “WormGPT” style tools, essentially LLM wrappers with the safety guardrails removed, at subscription prices of $200 to $1,700 per month.

The implications here are significant. AI lowers the barrier to entry for technically unsophisticated threat actors, increases the quality and scale of social engineering attacks, and makes it harder for defenders to distinguish malicious content from legitimate communications.

Narcotics and Controlled Substances

Dark web drug marketplaces remain one of the most active sectors of the underground economy. Our DARKSEARCH crawls in Q1 2026 revealed multiple operational marketplaces with extensive product catalogues, professional vendor storefronts, and established escrow systems. The sophistication of these operations is notable: vendor pages include lab-testing claims, customer reviews, volume discount tiers, and next-day delivery (NDD) options for domestic shipments.

Three marketplaces stood out during our research. TheBreakingBad, a dedicated vendor storefront operating with a full e-commerce style interface, offered a comprehensive catalogue of stimulants, opiates, and dissociatives with granular volume pricing. Abacus Market, a multi-vendor marketplace, carried similar inventory with slightly different pricing. Tor Market, which operates as a broader multi-category darknet marketplace (also listing firearms, documents, and hacking tools), hosted 47 drug products across multiple vendors at the time of our crawl.

Stimulants

Cocaine remains the most commonly listed stimulant. Colombian cocaine claiming 94%+ purity was available across multiple markets. Crystal methamphetamine was the second most prevalent stimulant listing, with a notably well-developed volume pricing structure from European vendors. Amphetamine paste, particularly popular on European markets, was available in both standard (74%) and premium (94%) purity grades.

ProductPrice RangeVolume PricingSource Market
Colombian Cocaine 94%+$50 – $80/gBulk from $35/g at 100g+Tor Market, Abacus
Crystal Meth 94% (Mexican)€10/g€80/10g, €700/100g, €5,500/kgTheBreakingBad
Crystal Meth (ICE)$99/10g$249/25g, $1,000/100gAbacus Market
Speed Amphetamine 94%€22/10g€90/100g, €700/kgTheBreakingBad
Speed Amphetamine 74%€10/10g€77/100g, €555/kgTheBreakingBad
3-MMC (Metaphedrone)€10/g€350/100g, €3,000/kgTheBreakingBad
MDMA Champagne 84%+$6.50 – $18/g$8/g at 250g bulkAbacus Market
XTC Pills 250mg MDMA€12.50/10 pills€80/100, €750/1,000 pillsTheBreakingBad
XTC Pills (240mg, various)$135 – $200/10 pillsMultiple brands availableTor Market

Opiates and Opioids

Heroin remained available from specialist vendors, with Iranian-sourced uncut product marketed as the premium option. The pricing structure on TheBreakingBad was particularly detailed, offering nine quantity tiers from a single gram to a full kilogram. This level of volume pricing suggests these vendors are servicing both individual users and mid-level distributors.

Prescription opioids also featured prominently. Oxycontin (40mg tablets) and Percocet (5/325mg) were listed on Tor Market, though exact per-unit pricing was often obscured behind “add to cart” interfaces that required account creation to view.

ProductPrice RangeVolume PricingSource Market
Heroin Uncut (Iranian)€22.50/g€175/10g, €1,600/100g, €13,500/kgTheBreakingBad
Heroin #3 (60-70%)$50 – $55/3gMid-grade, EU sourcedTor Market
Oxycontin 40mg (20ct)$120 – $200Prescription tabsTor Market
Percocet 5/325mg (70ct)$150 – $250Price per bottle est.Tor Market
Fentanyl patches/pills$50 – $150Limited listings (high risk)Various

Cannabis

Cannabis products dominated by volume of listings. UK-based vendors advertised next-day delivery (NDD) on multiple strains, essentially running a delivery service comparable to legitimate e-commerce. Listings included premium strains such as OG Cookies, Super Silver Haze, Gorilla Glue, and Amnesia Haze, with clear quantity tiers.

ProductPrice RangeVolume/NotesSource Market
Gorilla Glue (7g)£42 (~$53)UK NDD availableAbacus Market
OG Cookies (various)$50 – $120/quarterMultiple vendorsTor Market
Amnesia Haze (100g)$400 – $600Bulk listingAbacus Market
Super Silver Haze$35 – $80/quarterDutch sourcedTor Market
Cannabis (French market)Varies192 products listedFR marketplace

Psychedelics and Dissociatives

The psychedelics market showed strong activity, with psilocybin products packaged in consumer-friendly formats (chocolate edibles, microdose capsules) and ketamine available from multiple vendors. LSD pricing was harder to pin down through DARKSEARCH alone, but cross-referencing with forum discussions suggests typical street-equivalent pricing in the $5 to $15 per tab range.

ProductPrice RangeNotesSource Market
Psilocybin Chocolate (4g)$40Consumer-packaged edibleTor Market
Psilocybin Capsules 150mg (x100)$80 – $150Microdose formatTor Market
Ketamine S-Isomer€10/g€175/50g, €1,850/kgTheBreakingBad
LSD Tabs$5 – $15/tabForum pricing cross-refMultiple
XTC/MDMA (ecstasy, various)€12.50 – $200/10 pillsBrand-dependent pricingMultiple markets

Prescription Pharmaceuticals

Beyond controlled opioids, a range of prescription medications was available. Benzodiazepines (particularly Xanax and Rivotril) were listed at a fraction of pharmacy prices. Erectile dysfunction medications (Cialis) appeared as bulk listings, likely diverted or counterfeit product.

ProductPrice RangeNotesSource Market
Xanax 2mg (50 pills)€5 – €25Alprazolam, likely pressedAbacus Market
Rivotril 2mg (20 pills)$15 – $40ClonazepamTor Market
Cialis (50 tabs)$120 – $200Bulk pack, likely generic/counterfeitTor Market

The drug marketplace in 2026 functions like a professional retail operation. Escrow, customer reviews, volume discounts, refund policies, and domestic stealth shipping are standard. The operational maturity here mirrors what we have seen in the cyber services space, with vendor reputation systems driving quality competition.

Firearms and Ammunition

Firearms remain one of the most sensitive categories on the dark web. Our DARKSEARCH queries returned listings from multiple sources, including a dedicated storefront called “Gun and Shell Factory” and the firearms category on Tor Market (which carried 10 products at the time of our crawl). A vendor called “GlockZ” was also active with 7 listed products.

It is worth noting that firearms sales on the dark web carry the highest scam risk of any category. Law enforcement honeypot operations are well-documented in this space, and many “vendors” simply take payment and never deliver. That said, the listings themselves are informative for understanding what threat actors believe constitutes a reasonable market price, and the availability of these listings is itself a data point worth tracking.

Handguns

FirearmListed Price (USD)CalibreSource
Glock 17 Gen 4$4999mmTor Market
Glock 19$4509mmGun and Shell Factory
Glock 26$3509mmGun and Shell Factory
SIG Sauer P320$6009mmGun and Shell Factory
SIG Sauer P220$680 (sale from $800).45 ACPTor Market
Desert Eagle$899 (sale from $1,000).44 MagnumTor Market
Beretta M9$2499mmTor Market
Ed Brown Kobra$499.45 ACPGun and Shell Factory
CZ TS 2$8999mmGun and Shell Factory

Long Guns and Submachine Guns

FirearmListed Price (USD)TypeSource
AK-47$800 – $1,200Assault RifleGun and Shell Factory
AR-15$700 – $1,000Semi-Auto RifleGun and Shell Factory
UZI Pro$740Submachine GunGun and Shell Factory

Ammunition was also listed separately, though pricing data was less granular in our crawl results. The presence of both firearms and ammunition on the same marketplaces that sell drugs, stolen data, and hacking tools underscores the breadth of these platforms. Tor Market, for instance, carries categories for Counterfeits, Credit Card/CVV/Dumps, Documents, Drugs (47 products), Firearms and Ammo (10 products), Gadgets, and Hacking (13 products), all under one marketplace roof.

Compared to legitimate retail prices, dark web firearms are generally listed at a discount of 30% to 60% from retail, which reflects the risk premium inverted: buyers on the dark web are willing to pay less because of the high risk of scam, non-delivery, or law enforcement interception. From a threat intelligence perspective, the persistence of these listings indicates ongoing demand from individuals who cannot or will not purchase through legitimate channels.

Expanded Counterfeit and Fraud Services

Beyond the identity documents and financial instruments covered earlier, the dark web hosts a broader ecosystem of counterfeit goods and fraud services. Our expanded DARKSEARCH crawl revealed categories including counterfeit luxury goods, forged academic credentials, cryptocurrency fraud tools, and casino bonus exploitation kits.

Counterfeit Luxury Goods
Tor Market listed counterfeit luxury watches, with a Rolex Submariner Non-Date 41mm (model 124060) featured as a promoted product. Counterfeit luxury goods have historically been a smaller dark web category compared to clearnet operations, but their presence on multi-category darknet marketplaces suggests vendors are expanding their offerings to capture cross-selling opportunities from buyers already on the platform for other products.

Cryptocurrency Fraud Tools

Cryptocurrency fraud tools were among the most expensive single-item listings we encountered. The “Tor Amazon” marketplace (operating since 2019) offered an extensive catalogue including stolen Bitcoin wallets, fake USDT senders, wallet cracking tools, and compromised exchange accounts. The pricing here is particularly instructive.

ProductPrice (USD)DetailsSource
Stolen BTC Wallet (599 BTC)$42,000 – $59,900Priced at ~0.1% of wallet balanceTor Amazon
Atomic Wallet 1BTC+$4,000Pre-loaded compromised walletTor Amazon
Bitcoin Wallet w/ Seeds$1,500 – $6,500Wallet.dat with passphraseTor Amazon
Wallet.dat Passphrase Cracker$400Brute-force toolTor Amazon
Flash/Fake USDT Sender$300 – $500Spoofed transactionsTor Amazon
BTC Mnemonic Brute Tool£550 (~$690)12-phrase wallet crackerStandalone store
BTC Reverse Transaction Tool$400 – $600Transaction reversal exploitStandalone store
Leaked Data (16B accounts)$121,484Apple, Google, Binance etc.Tor Amazon

Financial Fraud and Money Movement

The money movement ecosystem on the dark web continues to grow. Services for laundering funds through compromised payment platforms, cloned cards, and bank transfer services were widely available. The Tor Amazon marketplace offered ATM-cloned cards with guaranteed balances, stolen Visa CC/CVV data, Binance account transfers, PayPal-to-Bitcoin conversion services, and casino bonus exploitation kits.

ProductPrice (USD)DetailsSource
ATM Cloned Card ($15K balance)$900Physical card, PIN includedTor Amazon
VISA CC/CVV ($8K balance)$400Virtual card with full detailsTor Amazon
PayPal $7K Verified Transfer$600 (sale from $700)Within 20 minutes worldwideTor Market
Visa Prepaid Clone ($7.5K)$630 (sale from $750)Physical + online usableTor Market
Binance Account Transfer$300 – $500BTC/ETH/USDT transfersTor Amazon
Paxful Accounts ($5.5K)$250 – $400Guaranteed balanceTor Amazon
Casino Bonus Exploit$150 – $350$3K-$10K bonus exploitationTor Amazon
Bank Flash SQR Tool$800 (sale from $1,500)Bank manipulation softwareTor Amazon
Aviator Predictor Hack AI$400Casino/gambling exploit toolTor Amazon
Gold Bars 100g (Pre-Owned)$8,500Physical delivery, escrowTor Amazon

Forged Documents and Credentials

Forged documents ranged from academic credentials (diplomas, degrees, professional certificates) to government-issued identity documents. Tor Market listed a dedicated Documents category with 4 products, while Tor Amazon carried a broader selection under their Documents department. The “USA Documents” product on Tor Amazon was rated 4.85 out of 5 and priced between $600 and $3,500, covering various forms of US identification.

ProductPrice Range (USD)NotesSource
USA Identity Documents (ID Card)$600 – $3,500Multiple ID types availableTor Amazon
Forged Diploma/Degree$200 – $800Various institutionsMultiple markets
Professional Certificates$150 – $500IT, medical, trade certsMultiple markets
Counterfeit COVID Certificates$50 – $150Declining demandForum listings
Deepfake Service$100 – $500Video/image manipulationTor directories

The breadth of these offerings paints a picture of a mature underground economy that mirrors, and in some ways parodies, legitimate commerce. Marketplaces offer escrow protection, customer reviews, vendor ratings, return policies, and even promotional sales events. Tor Amazon, for example, displays a running shopping cart total (one snapshot showed a cart worth $154,514), tracks “verified sellers” with sales counts, and runs sale pricing on multiple products. This operational maturity makes these platforms resilient and, from a threat intelligence perspective, worth continuous monitoring.

Pricing Comparison: 2022 vs 2026

The table below compares our 2022 findings with the current 2026 data across key categories. Prices are typical midpoint values.

Category2022 Average2026 AverageTrend
Credit card with CVV$243$15 – $40Decreased (oversupply)
Counterfeit currency (per $1K)$396$350 – $450Stable
DDoS service (monthly)$382$200 – $500Decreased (commoditised)
Residential proxy (monthly)$645$200 – $645Wider range, lower entry
Initial network access$7,700$2,000 – $5,000Decreased (median)
Ransomware kitN/A$500 – $5,000New category tracked
Crypto drainer kitN/A$200 – $1,000New category
Stealer log subscriptionN/A$100 – $1,024/moNew category
AI criminal toolsN/A$200 – $1,700/moNew category
Cocaine (per gram)$150 – $300$50 – $80Decreased (dark web discount)
Crystal Meth (per gram)$50 – $100$10 – $15Decreased significantly
Heroin (per gram)$100 – $200$22 – $55Decreased (direct sourcing)
Handgun (Glock 17)$1,500 – $2,500$450 – $500Decreased (high scam risk)
Stolen BTC WalletN/A$4,000 – $59,900New category
Forged ID Documents (US)$250 – $1,000$600 – $3,500Increased (quality premium)

The overarching trend is clear: established product categories have become cheaper as supply has increased, while new, more sophisticated offerings (RaaS, drainers, AI tools) have emerged at premium price points. The dark web economy is following the same pattern as legitimate tech markets, with commodity products racing to the bottom while innovation commands a premium.

Key Takeaways

The barrier to entry keeps falling. DDoS attacks for $10, phishing kits for $50, stolen accounts for a few dollars. The tools for cybercrime are cheaper and more accessible than ever. This has direct implications for the volume of attacks organisations should expect to face.

Stealer logs are the new oil. The stealer log economy has grown enormously. These logs, harvested from malware infections on individual machines, contain browser-saved passwords, session cookies, crypto wallet data, and more. They feed almost every other category: account takeover, initial access brokering, financial fraud, and identity theft.

Ransomware is a mature industry. With over 100 active groups tracked on our platform and well-established affiliate models, ransomware has moved from being an emerging threat to a structural feature of the threat landscape. The supply chain (IABs to RaaS operators to affiliates to money launderers) is well-oiled and efficient.

AI is an accelerant. While AI has not yet created fundamentally new attack types, it is making existing attacks more effective, more scalable, and more convincing. The appearance of AI-enabled tools as a distinct product category on the dark web is a development every security team should be tracking.

Crypto is the preferred battlefield. The emergence of crypto drainers as a major product category, combined with the growth in compromised exchange accounts, tells us that cryptocurrency users and platforms are now firmly in the crosshairs. The pseudonymous nature of crypto transactions makes this an attractive and growing target.

Drug marketplaces operate like professional retailers. The dark web drug economy has reached a level of operational maturity that mirrors legitimate e-commerce. Escrow, customer reviews, next-day delivery, volume discounts, lab-testing claims, and refund policies are now standard. Prices have dropped significantly compared to street equivalents, reflecting the efficiency of direct vendor-to-buyer models that bypass traditional distribution chains.

Firearms listings persist despite high scam risk. While firearms are consistently available on dark web marketplaces, this category carries the highest scam risk and is a known target for law enforcement honeypot operations. The listed prices (30% to 60% below retail) reflect this risk. The intelligence value here is less about the prices themselves and more about the persistent demand signal from individuals seeking to acquire weapons outside regulated channels.

The dark web is a one-stop shop. Multi-category marketplaces like Tor Market (drugs, firearms, counterfeits, hacking tools, documents, and financial fraud under one roof) and Tor Amazon (hacking, financial, electronics, documents, drugs, and guns) demonstrate that the underground economy has consolidated. A single marketplace visit can service everything from identity theft to substance procurement to weapon acquisition. This consolidation has implications for law enforcement, intelligence analysts, and risk modelling.

Don’t miss out!

Webinar: 2026 Dark Web Pricing Report

Conclusion

The dark web economy in 2026 is bigger, more diverse, and more sophisticated than it was in 2022. Prices for commodity products have dropped while new, higher-value categories have emerged. The professionalism of threat actors continues to increase, with customer support, affiliate programmes, and quality guarantees now standard across many marketplaces.

What has changed most since our 2022 report is the breadth. The dark web is no longer just a marketplace for stolen data and hacking tools. It is a fully integrated underground economy spanning narcotics, firearms, counterfeit goods, identity documents, cryptocurrency fraud, and digital services. Multi-category marketplaces have consolidated these offerings under single platforms, complete with escrow systems, vendor ratings, and promotional campaigns that would not look out of place on a legitimate e-commerce site.

For defenders and intelligence professionals, the key takeaway is that the cost of attacking your organisation, or acquiring the tools to do so, is low and getting lower. The investment needed to mount a credible phishing campaign, launch a DDoS attack, purchase a weapon, or obtain fraudulent identity documents is trivial compared to the potential payoff. This asymmetry is the fundamental challenge, and understanding the economics of the dark web is essential to building effective defences and informing policy.

At SOS Intelligence, we monitor these marketplaces continuously so our customers do not have to. Our DARKSEARCH platform indexes content across 50+ active dark web sources, and our analysts track emerging threats, new marketplace activity, and pricing trends in real time. If you want to understand what is being said about your organisation on the dark web, or if you need intelligence on any of the categories covered in this report, our platform gives you that visibility.

Passport photo by Kit (formerly ConvertKit) on Unsplash

Crypto Photo by Pierre Borthiry – Peiobty on Unsplash

Drugs photo by Colin Davis on Unsplash

Money hoto by Dmytro Glazunov on Unsplash

Gun photo by Tom Def on Unsplash

"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 20 April 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-25253

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25253

 


 

2. CVE-2026-25157

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25157

 


 

3. CVE-2026-24763

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-24763

 


 

4. CVE-2026-25475

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25475

 


 

5. CVE-2026-26322

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-26322

 


 

6. CVE-2026-27001

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-27001

 


 

7. CVE-2026-26329

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-26329

 


 

8. CVE-2025-43510

A memory corruption issue was addressed with improved memory handling. This issue is fixed in iOS 18.7.2 and iPadOS 18.7.2, iOS 26.1 and iPadOS 26.1, macOS Sequoia 15.7.2, macOS Sonoma 14.8.2, macOS Tahoe 26.1, tvOS 26.1, visionOS 26.1, watchOS 26.1. A malicious application may be able to cause unexpected system termination or write kernel memory.

https://nvd.nist.gov/vuln/detail/CVE-2025-43510

 


 

9. CVE-2026-20700

A memory corruption issue was addressed with improved memory handling. This issue is fixed in iOS 18.7.2 and iPadOS 18.7.2, iOS 26.1 and iPadOS 26.1, macOS Sequoia 15.7.2, macOS Sonoma 14.8.2, macOS Tahoe 26.1, tvOS 26.1, visionOS 26.1, watchOS 26.1. A malicious application may be able to cause unexpected system termination or write kernel memory.

https://nvd.nist.gov/vuln/detail/CVE-2026-20700

 


 

10. CVE-2025-43529

A memory corruption issue was addressed with improved memory handling. This issue is fixed in iOS 18.7.2 and iPadOS 18.7.2, iOS 26.1 and iPadOS 26.1, macOS Sequoia 15.7.2, macOS Sonoma 14.8.2, macOS Tahoe 26.1, tvOS 26.1, visionOS 26.1, watchOS 26.1. A malicious application may be able to cause unexpected system termination or write kernel memory.

https://nvd.nist.gov/vuln/detail/CVE-2025-43529

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 06 April 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-26329

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-26329

 


 

2. CVE-2026-26322

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-26322

 


 

3. CVE-2026-27001

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-27001

 


 

4. CVE-2026-25475

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-25475

 


 

5. CVE-2026-24763

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-24763

 


 

6. CVE-2026-25253

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-25253

 


 

7. CVE-2026-25157

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-25157

 


 

8. CVE-2024-38476

Vulnerability in core of Apache HTTP Server 2.4.59 and earlier are vulnerably to information disclosure, SSRF or local script execution via backend applications whose response headers are malicious or exploitable.

Users are recommended to upgrade to version 2.4.60, which fixes this issue.

https://nvd.nist.gov/vuln/detail/CVE-2024-38476

 


 

9. CVE-2025-43529

A memory corruption issue was addressed with improved lock state checking. This issue is fixed in iOS 18.7.2 and iPadOS 18.7.2, iOS 26.1 and iPadOS 26.1, macOS Sequoia 15.7.2, macOS Sonoma 14.8.2, macOS Tahoe 26.1, tvOS 26.1, visionOS 26.1, watchOS 26.1. A malicious application may cause unexpected changes in memory shared between processes.

https://nvd.nist.gov/vuln/detail/CVE-2025-43529

 


 

10. CVE-2025-43520

A memory corruption issue was addressed with improved lock state checking. This issue is fixed in iOS 18.7.2 and iPadOS 18.7.2, iOS 26.1 and iPadOS 26.1, macOS Sequoia 15.7.2, macOS Sonoma 14.8.2, macOS Tahoe 26.1, tvOS 26.1, visionOS 26.1, watchOS 26.1. A malicious application may cause unexpected changes in memory shared between processes.

https://nvd.nist.gov/vuln/detail/CVE-2025-43520

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 30 March 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-24763

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-24763

 


 

2. CVE-2026-25157

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-25157

 


 

3. CVE-2026-25253

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-25253

 


 

4. CVE-2026-27001

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-27001

 


 

5. CVE-2026-26329

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-26329

 


 

6. CVE-2026-26322

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-26322

 


 

7. CVE-2026-25475

OpenClaw is a personal AI assistant. Prior to version 2026.2.14, authenticated attackers can read arbitrary files from the Gateway host by supplying absolute paths or path traversal sequences to the browser tool’s `upload` action. The server passed these paths to Playwright’s `setInputFiles()` APIs without restricting them to a safe root. An attacker must reach the Gateway HTTP surface (or otherwise invoke the same browser control hook endpoints); present valid Gateway auth (bearer token / password), as required by the Gateway configuration (In common default setups, the Gateway binds to loopback and the onboarding wizard generates a gateway token even for loopback); and have the `browser` tool permitted by tool policy for the target session/context (and have browser support enabled). If an operator exposes the Gateway beyond loopback (LAN/tailnet/custom bind, reverse proxy, tunnels, etc.), the impact increases accordingly. Starting in version 2026.2.14, the upload paths are now confined to OpenClaw’s temp uploads root (`DEFAULT_UPLOAD_DIR`) and traversal/escape paths are rejected.

https://nvd.nist.gov/vuln/detail/CVE-2026-25475

 


 

8. CVE-2025-31277

The issue was addressed with improved memory handling. This issue is fixed in Safari 18.6, watchOS 11.6, visionOS 2.6, iOS 18.6 and iPadOS 18.6, macOS Sequoia 15.6, tvOS 18.6. Processing maliciously crafted web content may lead to memory corruption.

https://nvd.nist.gov/vuln/detail/CVE-2025-31277

 


 

9. CVE-2025-43520

The issue was addressed with improved memory handling. This issue is fixed in Safari 18.6, watchOS 11.6, visionOS 2.6, iOS 18.6 and iPadOS 18.6, macOS Sequoia 15.6, tvOS 18.6. Processing maliciously crafted web content may lead to memory corruption.

https://nvd.nist.gov/vuln/detail/CVE-2025-43520

 


 

10. CVE-2026-20700

The issue was addressed with improved memory handling. This issue is fixed in Safari 18.6, watchOS 11.6, visionOS 2.6, iOS 18.6 and iPadOS 18.6, macOS Sequoia 15.6, tvOS 18.6. Processing maliciously crafted web content may lead to memory corruption.

https://nvd.nist.gov/vuln/detail/CVE-2026-20700

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 23 March 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-32597

Exiv2 is a C++ library and a command-line utility to read, write, delete and modify Exif, IPTC, XMP and ICC image metadata. Prior to version 0.28.8, an uncaught exception was found in Exiv2. The vulnerability is in the preview component, which is only triggered when running Exiv2 with an extra command line argument, like -pp. Due to an integer overflow, the code attempts to create a huge std::vector, which causes Exiv2 to crash with an uncaught exception. This issue has been patched in version 0.28.8.

https://nvd.nist.gov/vuln/detail/CVE-2026-32597

 


 

2. CVE-2026-27631

Exiv2 is a C++ library and a command-line utility to read, write, delete and modify Exif, IPTC, XMP and ICC image metadata. Prior to version 0.28.8, an uncaught exception was found in Exiv2. The vulnerability is in the preview component, which is only triggered when running Exiv2 with an extra command line argument, like -pp. Due to an integer overflow, the code attempts to create a huge std::vector, which causes Exiv2 to crash with an uncaught exception. This issue has been patched in version 0.28.8.

https://nvd.nist.gov/vuln/detail/CVE-2026-27631

 


 

3. CVE-2025-61147

Exiv2 is a C++ library and a command-line utility to read, write, delete and modify Exif, IPTC, XMP and ICC image metadata. Prior to version 0.28.8, an uncaught exception was found in Exiv2. The vulnerability is in the preview component, which is only triggered when running Exiv2 with an extra command line argument, like -pp. Due to an integer overflow, the code attempts to create a huge std::vector, which causes Exiv2 to crash with an uncaught exception. This issue has been patched in version 0.28.8.

https://nvd.nist.gov/vuln/detail/CVE-2025-61147

 


 

4. CVE-2026-3606

Exiv2 is a C++ library and a command-line utility to read, write, delete and modify Exif, IPTC, XMP and ICC image metadata. Prior to version 0.28.8, an uncaught exception was found in Exiv2. The vulnerability is in the preview component, which is only triggered when running Exiv2 with an extra command line argument, like -pp. Due to an integer overflow, the code attempts to create a huge std::vector, which causes Exiv2 to crash with an uncaught exception. This issue has been patched in version 0.28.8.

https://nvd.nist.gov/vuln/detail/CVE-2026-3606

 


 

5. CVE-2026-25884

ImageMagick is free and open-source software used for editing and manipulating digital images. Prior to version 7.1.2-12, Magick fails to check for circular references between two MVGs, leading to a stack overflow. This is a DoS vulnerability, and any situation that allows reading the mvg file will be affected. Version 7.1.2-12 fixes the issue.

https://nvd.nist.gov/vuln/detail/CVE-2026-25884

 


 

6. CVE-2026-32239

Cap’n Proto is a data interchange format and capability-based RPC system. Prior to 1.4.0, when using Transfer-Encoding: chunked, if a chunk’s size parsed to a value of 2^64 or larger, it would be truncated to a 64-bit integer. In theory, this bug could enable HTTP request/response smuggling. This vulnerability is fixed in 1.4.0.

https://nvd.nist.gov/vuln/detail/CVE-2026-32239

 


 

7. CVE-2026-32240

Cap’n Proto is a data interchange format and capability-based RPC system. Prior to 1.4.0, when using Transfer-Encoding: chunked, if a chunk’s size parsed to a value of 2^64 or larger, it would be truncated to a 64-bit integer. In theory, this bug could enable HTTP request/response smuggling. This vulnerability is fixed in 1.4.0.

https://nvd.nist.gov/vuln/detail/CVE-2026-32240

 


 

8. CVE-2026-23925

VideoLAN VLC for Android prior to version 3.7.0 contains a path traversal vulnerability in the Remote Access Server routing for the authenticated endpoint GET /download. The file query parameter is concatenated into a filesystem path under the configured download directory without canonicalization or directory containment checks, allowing an authenticated attacker with network reachability to the Remote Access Server to request files outside the intended directory. The impact is bounded by the Android application sandbox and storage restrictions, typically limiting exposure to app-internal and app-specific external storage.

https://nvd.nist.gov/vuln/detail/CVE-2026-23925

 


 

9. CVE-2026-27596

ImageMagick is free and open-source software used for editing and manipulating digital images. Prior to version 7.1.2-12, Magick fails to check for circular references between two MVGs, leading to a stack overflow. This is a DoS vulnerability, and any situation that allows reading the mvg file will be affected. Version 7.1.2-12 fixes the issue.

https://nvd.nist.gov/vuln/detail/CVE-2026-27596

 


 

10. CVE-2026-31958

PyJWT is a JSON Web Token implementation in Python. Prior to 2.12.0, PyJWT does not validate the crit (Critical) Header Parameter defined in RFC 7515 §4.1.11. When a JWS token contains a crit array listing extensions that PyJWT does not understand, the library accepts the token instead of rejecting it. This violates the MUST requirement in the RFC. This vulnerability is fixed in 2.12.0.

https://nvd.nist.gov/vuln/detail/CVE-2026-31958

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 16 March 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-24763

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-24763

 


 

2. CVE-2026-25157

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25157

 


 

3. CVE-2026-25253

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25253

 


 

4. CVE-2026-25475

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-25475

 


 

5. CVE-2026-26322

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-26322

 


 

6. CVE-2026-26329

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-26329

 


 

7. CVE-2026-27001

OpenClaw is a personal AI assistant. Prior to version 2026.1.29, there is an OS command injection vulnerability via the Project Root Path in sshNodeCommand. The sshNodeCommand function constructed a shell script without properly escaping the user-supplied project path in an error message. When the cd command failed, the unescaped path was interpolated directly into an echo statement, allowing arbitrary command execution on the remote SSH host. The parseSSHTarget function did not validate that SSH target strings could not begin with a dash. An attacker-supplied target like -oProxyCommand=… would be interpreted as an SSH configuration flag rather than a hostname, allowing arbitrary command execution on the local machine. This issue has been patched in version 2026.1.29.

https://nvd.nist.gov/vuln/detail/CVE-2026-27001

 


 

8. CVE-2024-38476

Vulnerability in core of Apache HTTP Server 2.4.59 and earlier are vulnerably to information disclosure, SSRF or local script execution via backend applications whose response headers are malicious or exploitable.

Users are recommended to upgrade to version 2.4.60, which fixes this issue.

https://nvd.nist.gov/vuln/detail/CVE-2024-38476

 


 

9. CVE-2023-52658

In the Linux kernel, the following vulnerability has been resolved:

bpf: Reject narrower access to pointer ctx fields

The following BPF program, simplified from a syzkaller repro, causes a
kernel warning:

r0 = *(u8 *)(r1 + 169);
exit;

With pointer field sk being at offset 168 in __sk_buff. This access is
detected as a narrower read in bpf_skb_is_valid_access because it
doesn’t match offsetof(struct __sk_buff, sk). It is therefore allowed
and later proceeds to bpf_convert_ctx_access. Note that for the
“is_narrower_load” case in the convert_ctx_accesses(), the insn->off
is aligned, so the cnt may not be 0 because it matches the
offsetof(struct __sk_buff, sk) in the bpf_convert_ctx_access. However,
the target_size stays 0 and the verifier errors with a kernel warning:

verifier bug: error during ctx access conversion(1)

This patch fixes that to return a proper “invalid bpf_context access
off=X size=Y” error on the load instruction.

The same issue affects multiple other fields in context structures that
allow narrow access. Some other non-affected fields (for sk_msg,
sk_lookup, and sockopt) were also changed to use bpf_ctx_range_ptr for
consistency.

Note this syzkaller crash was reported in the “Closes” link below, which
used to be about a different bug, fixed in
commit fce7bd8e385a (“bpf/verifier: Handle BPF_LOAD_ACQ instructions
in insn_def_regno()”). Because syzbot somehow confused the two bugs,
the new crash and repro didn’t get reported to the mailing list.

https://nvd.nist.gov/vuln/detail/CVE-2023-52658

 


 

10. CVE-2023-53421

In the Linux kernel, the following vulnerability has been resolved:

bpf: Reject narrower access to pointer ctx fields

The following BPF program, simplified from a syzkaller repro, causes a
kernel warning:

r0 = *(u8 *)(r1 + 169);
exit;

With pointer field sk being at offset 168 in __sk_buff. This access is
detected as a narrower read in bpf_skb_is_valid_access because it
doesn’t match offsetof(struct __sk_buff, sk). It is therefore allowed
and later proceeds to bpf_convert_ctx_access. Note that for the
“is_narrower_load” case in the convert_ctx_accesses(), the insn->off
is aligned, so the cnt may not be 0 because it matches the
offsetof(struct __sk_buff, sk) in the bpf_convert_ctx_access. However,
the target_size stays 0 and the verifier errors with a kernel warning:

verifier bug: error during ctx access conversion(1)

This patch fixes that to return a proper “invalid bpf_context access
off=X size=Y” error on the load instruction.

The same issue affects multiple other fields in context structures that
allow narrow access. Some other non-affected fields (for sk_msg,
sk_lookup, and sockopt) were also changed to use bpf_ctx_range_ptr for
consistency.

Note this syzkaller crash was reported in the “Closes” link below, which
used to be about a different bug, fixed in
commit fce7bd8e385a (“bpf/verifier: Handle BPF_LOAD_ACQ instructions
in insn_def_regno()”). Because syzbot somehow confused the two bugs,
the new crash and repro didn’t get reported to the mailing list.

https://nvd.nist.gov/vuln/detail/CVE-2023-53421

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 09 March 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2025-0368

The Banner Garden Plugin for WordPress plugin through 0.1.3 does not sanitise and escape a parameter before outputting it back in the page, leading to a Reflected Cross-Site Scripting which could be used against high privilege users such as admin or unauthenticated users.

https://nvd.nist.gov/vuln/detail/CVE-2025-0368

 


 

2. CVE-2025-68613

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2025-68613

 


 

3. CVE-2016-10033

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2016-10033

 


 

4. CVE-2022-21227

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2022-21227

 


 

5. CVE-2021-40539

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2021-40539

 


 

6. CVE-2025-50708

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2025-50708

 


 

7. CVE-2024-21762

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2024-21762

 


 

8. CVE-2025-21204

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2025-21204

 


 

9. CVE-2024-34102

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2024-34102

 


 

10. CVE-2017-8759

Improper neutralization of special elements used in a command (‘command injection’) in Windows PowerShell allows an unauthorized attacker to execute code locally.

https://nvd.nist.gov/vuln/detail/CVE-2017-8759

 


"SOS
CVE Top 10

The SOS Intelligence CVE Chatter Weekly Top Ten – 02 March 2026

 

This weekly blog post is from via our unique intelligence collection pipelines. We are your eyes and ears online, including the Dark Web.

There are thousands of vulnerability discussions each week. SOS Intelligence gathers a list of the most discussed Common Vulnerabilities and Exposures (CVE) online for the previous week.

We make every effort to ensure the accuracy of the data presented. As this is an automated process some errors may creep in.

If you are feeling generous please do make us aware of anything you spot, feel free to follow us on Twitter @sosintel and DM us. Thank you!

 


 

1.  CVE-2026-2763

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2763

 


 

2. CVE-2026-2783

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2783

 


 

3. CVE-2026-2785

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2785

 


 

4. CVE-2026-2786

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2786

 


 

5. CVE-2026-2765

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2765

 


 

6. CVE-2026-2764

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2764

 


 

7. CVE-2026-2795

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2795

 


 

8. CVE-2026-2796

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2796

 


 

9. CVE-2026-2767

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2767

 


 

10. CVE-2026-2762

Race condition in the JavaScript: GC component. This vulnerability affects Firefox < 148 and Thunderbird < 148.

https://nvd.nist.gov/vuln/detail/CVE-2026-2762

 


1 2 3 4 32 33
Now recruiting MSSP partners · deploy dark web monitoring under your own brand in 48 hours | Sign up to the Partner Portal →
Privacy Settings
We use cookies to enhance your experience while using our website. If you are using our Services via a browser you can restrict, block or remove cookies through your web browser settings. We also use content and scripts from third parties that may use tracking technologies. You can selectively provide your consent below to allow such third party embeds. For complete information about the cookies we use, data we collect and how we process them, please check our Privacy Policy
Youtube
Consent to display content from - Youtube
Vimeo
Consent to display content from - Vimeo
Google Maps
Consent to display content from - Google
Spotify
Consent to display content from - Spotify
Sound Cloud
Consent to display content from - Sound