fish-live-in-trees

Fish Live in Trees Production LLM Runtime Alignment Context Injection RACI Author: Sean Kavanagh Date: 2026-02-16 Environment: Gemini 3 Flash Free Tier - Public Production Interface Exploit Type: Contextual Pivot Factual Accuracy → Interpersonal Authenticity No jailbreak payloads. No special tool...

CLEANSTART-2024-CW18426 When browsing a document hosted on an IP address, an attacker could insert certain characters to flip domain and path information in the address bar

Security vulnerability affects the librewolf package. When browsing a document hosted on an IP address, an attacker could insert certain characters to flip domain and path information in the address bar...

CacheTrap

CacheTrap Repository for "CacheTrap: Unveiling a Stealthier Gray-Box Trojan against LLMs" , IEEE/ACM International Conference on Computer-Aided Design ICCAD, 2026. arXiv CacheTrap searches for a single bit flip in the KV cache of a fine-tuned LLM classifier and measures the resulting attack succe...

CLEANSTART-2024-AF69663 When browsing a document hosted on an IP address, an attacker could insert certain characters to flip domain and path information in the address bar

Security vulnerability affects the firefox package. When browsing a document hosted on an IP address, an attacker could insert certain characters to flip domain and path information in the address bar...

PT-2026-133653

Blender before 4.5.15 and 5.x before 5.2.3 contains a heap buffer overflow vulnerability in the blend file thumbnail extractor caused by 32-bit integer overflow in thumb data vertical flip. Attackers can craft a .blend file with malicious TEST block dimensions that trigger an out-of-bounds write...

Amplifying Latency and Cost of LLM Inference Via Rowhammer Attack

Large language models LLMs achieve high inference efficiency using optimized GPU kernels generated by just-in-time JIT compilation frameworks such as Triton. Triton compiles GPU kernels through an MLIR/LLVM-based pipeline. We investigate whether a Rowhammer-induced bit-flip in the CPU's DRAM can...

Astra Linux – vulnerability in linux, linux-5.10, linux-5.15, linux-6.1

In the Linux kernel, the following vulnerability has been resolved: drm: Do not unref the same fb multiple times by mistake due to deadlock handling If a deadlock occurs after the fb lookup in drmmodepageflipioctl, we proceed to unref the fb and then retry the entire operation from the beginning...

Astra Linux – vulnerability in libstb

StBImage is a single-file MIT-licensed library for image processing. If stbiloadgifmain within stbiloadgiffrommemory fails, it returns a null pointer and may leave the z variable uninitialized. If the caller also sets the vertical flip flag, the function proceeds and calls stbiverticalflipslices...

Astra Linux – Vulnerability in libstb

stbimage is a single-file library licensed under MIT that is used for processing images. When stbisetflipverticallyonload is set to TRUE, and reqcomp is set to a value that does not match the actual number of components per pixel, the library attempts to flip the image vertically. A specially...

Astra Linux – Vulnerabilities in Linux 5.10, Linux 5.15, Linux 6.1

In the Linux kernel, the following vulnerability has been resolved: drm/mediatek: Fixed a null pointer crash in mtkdrmcrtcfinishpageflip. It is possible that mtkcrtc-event is NULL in mtkdrmcrtcfinishpageflip. The pendingneedsvblank value is set by mtkcrtc-event, but in mtkdrmcrtcatomicFlush, it...

Astra Linux – vulnerability in linux, linux-5.10

In the Linux kernel, the following vulnerability has been resolved: net/tls: Fix flipped sign in tlserrabort calls sk-skerr appears to expect a positive value; this is a convention ktls does not always follow, leading to memory corruption in other code. For example, kworker tlsencryptdone..., err...

Astra Linux – Vulnerability found in Linux 5.10, Linux 5.15

In the Linux kernel, the following vulnerabilities have been resolved: drm/gma500: Fixed a bug where a sleeping function called from an invalid context caused errors. The function gmacrtcpageflip held the eventlock spinlock while calling crtcfuncs-modesetbase, which requires the wwmutex. The only...

Space over Statistic, Retraining over Ranking: Auditing Training Labels under Label-Only Backdoors

Some backdoor attacks never touch an image. FLIP only changes labels: on CIFAR-10, flipping 3% of them makes a ResNet-32 send 99.7% of triggered source-class images to the attacker's class. The usual defenses look for the poison inside a model trained on those same labels. So when they struggle, ...

ceph: ceph: RGW STS session tokens vulnerable to CBC bit-flip attack enabling admin privilege escalation

A flaw was found in Ceph RGW's STS Security Token Service session token implementation. The STS tokens use the same unauthenticated AES-128-CBC encryption as CephX, which lacks message authentication and uses a hardcoded initialization vector. Because there is no integrity protection on the token...

BARE-AI: Bit-Flip Attack Resilience in AI Hardware through Built-In Performance Monitors

Deep Neural Networks DNNs are integral to many safety critical systems, yet they remain highly vulnerable to bit-flip attacks BFAs, where a few memory level perturbations can drastically degrade accuracy. Existing defenses incur significant hardware overhead, depend on retraining, or fail against...

PT-2026-106648

Name of the Vulnerable Software and Affected Versions Linux kernel versions prior to 7.2.10-1.1 Description A memory leak occurs in the Direct Rendering Manager DRM when an out fence ptr is provided without the DRM MODE PAGE FLIP EVENT being set. In this scenario, a drm pending vblank event is...

Online AutoML: Evaluating Poisoning Attacks on Adversarial Training Defense Strategy in IoT Networks

Machine learning ML-powered poisoning attack vectors are adversarial maneuvers whereby an attacker intentionally inserts, corrupts, or alters training data to distort an ML model's learning process. The objective is to diminish model efficacy, instill biases, induce misclassifications, or include...

Security-Enhanced Seed-Based Weight Quantization for Large Language Models

Large language models LLMs incur substantial storage, memory-bandwidth and energy costs, motivating compact weight representations. Existing seed-based compression methods reconstruct weights from compact pseudo-random representations but do not explicitly account for the non-uniform sensitivity ...

ALIBI: Adversarial Legitimacy Injection in Binary Input against LLM Malware Analyzers

Large language models are being integrated into malware triage workflows as reasoning components that summarize static evidence and produce analyst-facing verdicts. This paper shows that the same reasoning capability introduces a new attack surface. We present ALIBI, a semantic cover story attack...

PT-2026-93457

Name of the Vulnerable Software and Affected Versions Linux kernel affected versions not specified Description A memory leak occurs in the Direct Rendering Manager DRM when PAGE FLIP EVENT is used and a signal is sent to the thread. Specifically, if an event is allocated in the prepare signaling...