Malicious themes steal iPhone crypto wallets

Hi there, this is your daily ☕️ Cyberpresso.

In today's Cyberpresso:

📱 Malicious themes steal iPhone crypto wallets

💊 McKesson breached via phishing

🔓 Exploit targets Microsoft Exchange servers

📌 Token exploit drains $75M from lender

🌐 Hijacked routing delivers malicious update

Plus: 💡 6 strategies & tactics, 🎁 7 other news you might like, 🧰 6 tools, and 📚 5 papers.

📱 Malicious themes steal iPhone crypto wallets LINK

Researchers at Socket found 13 malicious Composer theme packages on Packagist, published across five vendor namespaces, that inject JavaScript into Vietnamese movie and comic streaming sites to steal crypto wallets from iPhone visitors.

On iPhones, the injected loader pulls a FUNNULL-hosted WebKit-to-kernel chain that weaponizes two public, patched WebKit bugs (CVE-2025-31277 and CVE-2025-43529, the latter confirmed by Apple as exploited in a targeted attack) against devices running iOS 18.4 through 18.6.x that have not updated.

After a GPU-process pivot and a kernel escape, already fixed in iOS and macOS 26.1, making it a known unpatched bug used against outdated devices, an August redeployed payload steals keychain wallet seeds and mnemonics from Bitget, BitKeep, Bitpie, Phantom, Tonkeeper, Trust Wallet, and OKX.

💊 McKesson breached via phishing LINK

McKesson, a major U.S. healthcare and pharmaceutical distributor, has disclosed a breach in which the extortion group ShinyHunters claims to have stolen roughly 284 million patient records after gaining unauthorized access to third-party applications.

Attackers reportedly started with voice phishing (vishing) calls to employees to compromise Okta single sign-on accounts, then used those credentials to reach McKesson's Salesforce and Snowflake environments, taking about 1 terabyte of data over four days.

The stolen data allegedly includes names, addresses, dates of birth, Social Security numbers, and medical record numbers; McKesson discovered the incident on August 25, 2026, confirmed the theft, warned of possible intermittent service degradation, and has not yet determined materiality.

🔓 Exploit targets Microsoft Exchange servers LINK

Public exploit code now targets on-premises Microsoft Exchange servers through CVE-2026-62911, an authentication-bypass flaw disclosed after Pwn2Own Berlin 2026, where researcher Orange Tsai chained it to gain SYSTEM-level code execution for a $200,000 reward.

A proof-of-concept is now circulating, though defenders should treat it as unverified; the bug alone is an elevation-of-privilege issue and only reaches full remote code execution when chained with relay, authorization, and file-write weaknesses.

The published method coerces NTLM authentication from one Exchange server and relays it to another's MRSProxy service, which lacks Extended Protection for Authentication, to write an ASPX web shell and run commands as SYSTEM; apply Microsoft's August 2026 Exchange updates.

📌 Token exploit drains $75M from lender LINK

Cronos Network shut down its entire blockchain on Sunday after attackers exploited Tectonic, its largest lending protocol, manipulating a token's price to drain roughly $75 million in preliminary losses.

The attacker pushed TONIC, Tectonic's thinly traded governance token, about 100-fold higher within roughly 20 minutes, then deposited the inflated position as collateral, its 20% collateral factor let the same tokens back far larger loans while the manipulated price held.

Only about $6 million reached Ethereum before validators, capped at 100 under the network's proof-of-authority setup, coordinated an emergency halt that trapped most funds on-chain; Crypto.com said its app and exchange were unaffected and Tectonic told users not to interact with the protocol.

🌐 Hijacked routing delivers malicious update LINK

Attackers hijacked internet routing for Softaculous's Virtualizor update servers between August 28 and August 30, redirecting traffic to their own systems and pushing a malicious update package to a small number of servers.

The attackers made an unauthorized BGP announcement for the 162.55.80.0/24 range from AS62390 (NexonHost) via transit provider Zet.net, advertising a more-specific route than Hetzner's legitimate one so networks preferred it while keeping Hetzner's origin number to look inconspicuous.

Because certificate-validation traffic was also diverted, the attackers obtained a valid Let's Encrypt TLS certificate for virtualizor.com and related domains, so clients saw no warnings; routing is now restored, and administrators should review update activity, package integrity, and logs from the affected period.

💡 Strategies & Tactics

> Breaking the Seal: Static Deobfuscation of JSCeal’s Compiled V8 Bytecode: Check Point built a free static tool that unpacks JSCeal's compiled, obfuscated cryptocurrency-stealer code so analysts can read its logic without running the malware.

> OEMpocalypse Now: Targeting the custom Android layers phone makers like Samsung and Xiaomi add lets one root exploit span an entire brand's lineup regardless of chipset.

> How to Build a Hypothesis-Driven Threat Hunting Program: Build threat hunting as a documented program with structured hypotheses and telemetry checks so results stay consistent, prove coverage gaps, and survive staff turnover.

> Automate IAM Identity Center governance with continuous discovery and reporting: Explains how to track who can access which AWS applications by deploying a tool that continuously discovers user assignments and generates audit-ready reports.

> AWS S3 Bucket Security: Find the Secrets Hiding Outside Git: Scan S3 storage buckets for exposed credentials, since 28% of secret leaks now hide in logs and backups rather than code repositories.

> This Windows tool lets you see exactly which apps are phoning home (and lets you cut them off): Portmaster, a free tool, shows every outbound connection your Windows apps make and lets you block trackers or telemetry per app.

Other news you might like

• Hackers Launch Password Spraying Attacks Against AWS Root Accounts at 150+ OrganizationsLINK

• Mirage Kitten targeting aviation and FinTech sectors across the Middle East and Africa with a new malware setLINK

• Financially Motivated Threat Actor BREEZE COMET Targets BrazilLINK

• Nightmare Eclipse Drops ‘HardBreacher’ Kaspersky Product ExploitLINK

• New RevStealer malware spreads as fake Claude Opus 5 desktop appLINK

• Infostealers are hijacking Claude accounts at users’ expenseLINK

• Indeed Job Scam: Fake Recruiters Are Spreading Android SpywareLINK

🧰 Trending tools

Kastra: runtime authorization layer that enforces policies on AI agent actions before execution, blocking unauthorized tool use, prompt injection, and data exposure with sub-millisecond latency.LINK

Perfai Security: autonomously tests AI-built apps for access-control, business-logic, and prompt-injection flaws, then opens pull requests with confirmed fixes around the clock.LINK

qsa.sh: runs external port and vulnerability scans of your public IP using naabu, nmap, and nuclei, streaming results to your terminal in seconds with no install.LINK

HOL Guard: a local firewall for AI agents that intercepts risky actions like deleting production data or exposing secrets before they runLINK

Cynative Security Research Agent: open-source AI CLI that answers plain-language security questions across code, cloud, and runtime with read-only, IAM-enforced access to production.LINK

Aegisora: an open-source proxy layer that guards LLM agents with least-privilege API access, PII masking, prompt-injection blocking, and audit logging.LINK

📚 Trending papers & reports

Virtual reality logins can use passwords tied to the surrounding 3D scene instead of digits, and in a 66-person study over 2-3 weeks people remembered them better and found the process less mentally taxing than 6-digit PINs.LINK

Video watermarking lets anyone verify a clip's origin with just a public key while making it impossible to forge, proving authentic on 99.3% of 1000 real clips and never accepting a fake key.LINK

Encryption-key extraction shows that a simple, standard AI setup can steal a chip's full secret key from raw power measurements, matching specialized tools while training in under 3.3 hours on one graphics card.LINK

AI agent accountability gives a way to prove which automated service produced a given output and whether it was authorized to hand off work, catching forged claims that older single-signature methods miss.LINK

Backdoored language models can be cleaned by pinpointing the specific hidden triggers that make them misbehave, cutting successful attacks to under ~11% for one attack type and ~15% for another without hurting normal performance.LINK

See you tomorrow for a new dose of ☕️ Cyberpresso!