Results for “Attacks”

arXiv (research papers) Lukáš Brůna SI

Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs

Large Language Models (LLMs) consume and produce a single sequence of text; hence, if text can be added to the beginning of the LLM's response, i.e., an output prefix, then all subsequent tokens will be conditioned on it. This output-prefix attack technique is a cheap black-box prompt injection. Prior work has shown this type of attack can reliably jailbreak non-reasoning models. Most reasoning models add an intermediate scratchpad reasoning step before the assistant's final…

Read at arXiv (research papers)

arXiv (research papers) Karina Elzer SI

OllamaDrama: Designing and Deploying a Honeypot to Measure Attacks on Exposed LLM Infrastructure

Publicly exposed large language model (LLM) infrastructure creates a growing attack surface, yet real-world targeting remains poorly understood. We present Ollure, a low- and medium-interaction honeypot that emulates the Ollama API without a backend LLM. Spanning four deployments across cloud and university networks, Ollure operated for 84 days and recorded 290,887 interactions from 2,793 unique source IP addresses. Most of the activity consisted of automated discovery…

Read at arXiv (research papers)

Hacker News northbridgedev SI

Can open-source prompt-injection detectors catch realistic AI agent attacks?

Article URL: https://github.com/rudratoshs/buried-injections Comments URL: https://news.ycombinator.com/item?id=49827019 Points: 8 # Comments: 4

Read at Hacker News

arXiv (research papers) Tiantong Wu SI

Decision Hijacking: Prompt Injection Attacks on Jev's Typed Probabilistic Decisions

Most studies of prompt injection focus on generative agents, leaving their effects on models with schema-defined outputs unclear. We examine these effects in Jev, a non-generative decision model, using 510 reconstructed InjecAgent cases. Malicious content shifts action probabilities but rarely causes Jev to select the attacker's target. Override markers reduce this influence, while claims of contextual relatedness have small effects. Adaptive attacks using score feedback…

Read at arXiv (research papers)

Africa at LSE blog Mark Briggs Geopolitics

25 years after 9/11 West Africa bears the brunt of the War on Terror

Instead of eradicating terrorism, 25 years of the War on Terror have left parts of West Africa suffering from increased securitisation and more terrorist attacks, writes Hakeem Onapajo. Twenty-five years ago, the terror attacks on the World Trade Center and the Pentagon sent shockwaves across the world. The attack, by Osama bin Laden’s al-Qaeda, marked … Continued The post 25 years after 9/11 West Africa bears the brunt of the War on Terror first appeared on Africa at LSE.

Read at Africa at LSE blog

Committee to Protect Journalists CPJ Staff Geopolitics

Assaulted and threatened: risks surge for journalists ahead of Serbia elections

Berlin, September 22, 2026—As Serbians prepare to vote in October parliamentary elections, journalists covering the polls are at serious risk of threats, violence and hostility of the sort that has become increasingly common during President Aleksandar Vučić’s time in office. Recent attacks on reporters covering political rallies ahead of the vote on October 25 provide an...

Read at Committee to Protect Journalists

Journal articles (OpenAlex) Mengxiao Chang, Mengxiao Chang, Kun Wang et al. · Array Gadgets

A lightweight secure communication protocol in smart homes

Resource-constrained smart home devices usually suffer from limited computation capability, constrained memory, unstable links, and exposure to eavesdropping, replay, and man-in-the-middle attacks. To address these issues, we propose a lightweight secure communication protocol for smart home environments. At the core of the protocol is KeyGen, a lightweight session key derivation and state evolution mechanism that jointly binds a pre-shared key, negotiated nonces, and an…

Read at Journal articles (OpenAlex)

arXiv (research papers) Fernando Outeda SI

Decoding Guardrails: XAI-Guided Perturbation Analysis of Prompt Injection Detection

Large language models (LLMs) are increasingly deployed in production systems, raising concerns about their exposure to adversarial manipulation through prompt injection and jailbreak attacks. Classifier-based guardrails, such as Prompt Guard 2, are widely used as a first line of defense against such attacks, but their internal decision logic is largely opaque to both defenders and attackers. This paper presents an exploratory case study that applies explainable artificial…

Read at arXiv (research papers)

Committee to Protect Journalists CPJ Staff

Journalists attacked while covering violent farmers’ protest in Romania

Berlin, September 21, 2026—Romanian authorities must swiftly investigate attacks on journalists covering last week’s farmers’ protest in Bucharest, and make sure those responsible are held accountable while also respecting journalists’ right to protect their sources, the Committee to Protect Journalists said Monday. After the protests on September 15, police publicly requested footage from media, civil...

Read at Committee to Protect Journalists

Journal articles (OpenAlex) Sana Chavoshian, Hilal Alkan · Anthropology and Medicine Geopolitics

Trafficking harm: extractivism and tree affordances in the southwestern borderlands of Iran

In Khuzestan, Iran, Conocarpus erectus, a nonendemic, fast-growing evergreen tree species with rich foliage has been widely planted to fight the consequences of extractivism, war, and climate change. The trees were supposed to form a physical barrier to the dust-laden winds while releasing oxygen to alleviate air pollution. Yet the roots of the tree destroyed infrastructure and its pollen proved even more hazardous, provoking asthma attacks in large parts of the population…

Read at Journal articles (OpenAlex)

Article 19 kivilcimbuken

Türkiye: End attacks on LGBTQ+ organisations, activists and protesters now

We, the undersigned organisations, condemn the large-scale police raids and detentions targeting LGBTQ+ activists and human rights defenders, organisations, and community spaces across Türkiye. We call for the immediate protection and release of those detained, and respect for their procedural rights. We also call for an end to any measures obstructing the fundamental rights to

Read at Article 19

The Guardian Mali Jason Burke International security correspondent

Al-Qaida and IS-linked violence in Africa’s Sahel belt set to hit record levels this year

Twenty-five years after 9/11, figures compiled by conflict monitor underline how Sahel region has become centre of Islamist terror Islamist extremist violence in Africa’s Sahel region is set to reach historic highs after a series of sweeping offensives by groups linked to al-Qaida and Islamic State, according to new figures compiled by the conflict monitor Acled. As the US marks the 25th anniversary of the 9/11 attacks launched by al-Qaida against targets in Washington and…

Read at The Guardian Mali

Journal articles (OpenAlex) Saleh M. Altowaijri · Mathematics Gadgets

A Zero-Trust Micro-Segmentation Framework with Federated Anomaly Detection for Heterogeneous Smart Home Ecosystems

Today, with the emergence of the Internet of Things (IoT) in smart home technology, new security issues have arisen, including device diversity, data privacy concerns, and susceptibility to advanced, complex attacks. Traditional Intrusion Detection Systems (IDS) face critical challenges, including single points of failure, high communication overhead, and inherent privacy violations, which the proposed decentralised approach addresses. In this paper, a novel Federated…

Read at Journal articles (OpenAlex)

arXiv (research papers) Xiaoyan Li SI

Universal Defenses for Tool-Integrated LLM Agents Against Adversarial Attacks

Large Language Model (LLM) agents have demonstrated impressive capabilities across a variety of domains, particularly when integrated with external tools for multi-step task completion. However, they are increasingly vulnerable to adversarial attacks, including direct prompt injection, indirect prompt injection, memory poisoning, and backdoor attacks, which exploit the model's openness to prompt injection and tool manipulation. In this work, we explore practical and…

Read at arXiv (research papers)

arXiv (research papers) Bingzheng Wang SI

ActGuard: Pre-execution Action Auditing against Indirect Prompt Injection in LLM Agents

Large language model (LLM) agents interact with external environments through tool invocation, but tool outputs can also expose them to indirect prompt injection (IPI) attacks. Existing defenses mainly rely on prompt hardening, content filtering, pre-generated plans, or permission constraints. These approaches often struggle with complex tasks or over-sanitize external content, making it difficult to balance security and utility. The key challenge is therefore to preserve…

Read at arXiv (research papers)

arXiv (research papers) Donato Mecca SI

SkillSecurer: Detecting and Patching Prompt-Injection Vulnerabilities in AI Agent Skills

Agent skills extend AI agents with reusable instructions, scripts, and configuration, but are also open to new attacks to influence an agent's decisions and actions. To address these risks, we present SkillSecurer, a fully agentic framework for generating, detecting, localising, and remediating security risks in agent skills. Its red agent generates context-compatible injections across nine threat types while recording the exact modification; its blue agent analyses complete…

Read at arXiv (research papers)

The Guardian Afghanistan Associated Press Geopolitics

CIA releases dozens of presidential daily briefs on Bin Laden and al-Qaida from before 9/11

The presidential briefing documents were declassified on the 25th anniversary of the September 11 attacks Several dozen presidential briefing documents declassified on Friday, the 25th anniversary of the September 11 attacks, detail intelligence that government officials had gathered about Osama bin Laden and al-Qaida leading up to the hijackings that killed nearly 3,000 people and upended American national security. The documents illustrate security analysts’ tracking of…

Read at The Guardian Afghanistan

The Guardian Afghanistan Shadi Khan Saif

After 9/11, I lived through Australia’s long war in Afghanistan. There are lessons the world must learn | Shadi Khan Saif

For Australians, Afghanistan became the country where a generation of soldiers served their longest war. For Afghans, it became a country where another generation struggled to build a new life A quarter of a century ago today, the world witnessed a devastating series of attacks on the United States that would change the lives of millions across the globe, setting off a vicious cycle of violence that continues to reverberate. I vividly remember watching the breaking news of…

Read at The Guardian Afghanistan

arXiv (research papers) Viet K. Nguyen SI

An Experimental Evaluation of Multimodal Prompt Injection Attacks on Agentic AI Frameworks

Agentic AI frameworks let a language model plan, keep memory, and call tools that reach real files, mail, and services. Most of these agents also read images, which gives an attacker a way to put text into the agent's context without going through the user. We present MMPIBench, a reproducible benchmark that measures what happens next. It delivers a fixed set of attacks through six visual carriers (OCR text, overlays, EXIF metadata, QR codes, fake interfaces, and hybrids)…

Read at arXiv (research papers)

arXiv (research papers) Boyang Zhang SI

CoER: Defending against Adaptive Indirect Prompt Injection via Adversarial Co-Evolution and Refinement

Language-model agents are vulnerable to indirect prompt injection (IPI) during tool use: adversarial instructions hidden in untrusted tool outputs can covertly redirect legitimate task execution. Existing work often trains and evaluates defenses against fixed attacks that do not adapt to the defender's behavior, so the resulting defenses may struggle against adaptive attacks in real-world settings. We argue that a strong defense against adaptive IPI must adapt during…

Read at arXiv (research papers)

arXiv (research papers) Asif Pinjari SI

AgentDrift: A Step-Labeled Benchmark of Injection-Hijacked LLM Agent Trajectories

LLM agents complete tasks by issuing sequences of tool calls, and every observation they read is a channel through which an indirect prompt injection can enter. A successful injection has a characteristic shape when the trajectory is read in order: a benign prefix gives way to actions that serve the attacker rather than the user. Existing benchmarks measure whether such attacks succeed against live agents, and existing guard models judge a trace as a whole; no public corpus…

Read at arXiv (research papers)

The Guardian Belarus Kate Connolly in Berlin, Lorenzo Tondo and Pjotr Sauer

Germany urges EU to cut Russians’ holiday visas after sabotage attacks

Belarusian said to have had tourist visa issued by Italy is investigated over alleged attempt to attack Leipzig airport Germany is urging its European partners to introduce tighter limitations on Russian citizens entering the EU on tourist visas, after recent sabotage attacks which Berlin has blamed on Moscow. German authorities are investigating a Belarusian national, who also holds a Russian passport, over the alleged attempt to attack Leipzig airport using drones and…

Read at The Guardian Belarus

arXiv (research papers) Nivedita Singh SI

Shifting from Injection to Interaction: Rethinking Web Security in the Age of LLMs and Beyond

Large language models (LLMs) are becoming integral to web applications and browser agents, transforming online interactions while introducing new attack vectors and reshaping longstanding web vulnerabilities. Classical threats such as cross-site scripting (XSS) can be amplified through LLM-mediated interactions, while LLM-specific vulnerabilities can propagate across web applications, introducing attacks such as prompt injection. Securing modern web systems therefore…

Read at arXiv (research papers)

arXiv (research papers) Qikai Wang University of Electronic Science and Technology of China SI

Implicit Manipulation for Skill Selection in LLM Agents with Semantic Matching

Skill selection is a key stage in LLM-agent workflows, determining which installed skill should handle a user request. Existing attacks on this stage primarily rely on explicit prompt injection or instruction-level steering, which can expose recognizable manipulation signals. In this work, we identify a new implicit attack surface for skill selection: even when the user prompt and skill description appear benign in isolation, their semantic relationship can still be…

Read at arXiv (research papers)

Journal articles (OpenAlex) Christian S. Czymara · Social Sciences & Humanities Open Geopolitics

Public responses to jihadist terrorism on social media

In recent years, several major terror attacks in the name of political Islam have shaken Europe. This study analyzes public responses after such attacks to explore their impact on public views of immigration and intergroup relations. In particular, the study examines how public reactions to Jihadist terrorism are expressed on social media. The theoretical predictions, derived from combining terror management theory, group threat theory, and the concept of social resilience…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Qingsong Zou, Zhiyao Xu, Jingyu Xiao et al. · IEEE Transactions on Mobile Computing Gadgets

HomeGuardian: Detecting Behavioral Anomalies Through Implicit Profiling in Smart Home Systems

Smart homes, enabled by the Internet of Things (IoT), offer substantial convenience but also introduce security challenges. These often stem from anomalous behaviors, including improper user interactions and malicious attacks. Although various behavior-modeling methods have been proposed, userdevice interactions are often unobservable in practice due to encrypted traffic and limited vendor access, while anomaly detection models suffer severe degradation during cold start…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Ghada Abdelhady, Karim Wael Hussein, Islam Anwar Ali Gad · Scientific Reports Gadgets

Privacy-preserving intrusion detection in IoT smart homes using a federated hybrid 1D-CNN–LSTM model with explainable AI

The proliferation of Internet of Things (IoT) devices in smart home environments has dramatically expanded the attack surface for cyber threats, particularly botnet-driven Distributed Denial of Service (DDoS) attacks. Centralized Intrusion Detection Systems (IDS) are ill-suited to this domain because they violate user privacy, introduce single points of failure, and incur prohibitive communication overhead. Federated Learning (FL) offers a compelling privacy-preserving…

Read at Journal articles (OpenAlex)

Hacker News joshua_s_penman SI

Show HN: Semantic Overlays – an NX bit for LLM prompt injection (live demo)

I've built a new method for steering LLMs called Semantic Overlays, small trained adapters on a frozen model which change how it perceives a piece of its context. The most readily applicable usage is to mitigate prompt injection, and it lets us take a very-injectable Qwen-3.5-9B to SOTA scores on all the prompt injection benchmarks I could find. (They are only blackbox attacks, but I did NOT train on anything like them — whitebox attacks are out of scope for this paper) I'm…

Read at Hacker News

Journal articles (OpenAlex) Schirin Amir‐Moazami · Genealogy Geopolitics

The Political Theology of Germany’s Staatsräson: Exception, Enmity, and the Exposure of the Liberal Order

Germany’s unconditional support of Israel is encapsulated in the dictum “Israel’s security is Germany’s raison d’état” (Staatsräson). Shaped after World War II and formally articulated by Angela Merkel in her speech at the Knesset in 2008, the confession to Staatsräson has been intensified amid Israel’s ongoing genocide in Gaza, following the Hamas attacks on 7 October 2023. This article conceptualizes the consolidation of this extra- and paralegal doctrine through the lens…

Read at Journal articles (OpenAlex)

arXiv (research papers) Shiqian Zhao SI

ECLIPSE: Self-Evolving Stealthy Prompt Injection Attack against Long-Horizon Agentic Systems

Recently, large language model (LLM) agents, such as Codex, Claude Code, and OpenClaw, have become capable of planning and executing long-horizon tasks through repeated tool calls. This capability also creates new opportunities for prompt injection. Existing attacks either place the malicious objective in one explicit instruction, making it easy to detect, or distribute the intent across multiple execution stages, making successful completion unreliable. In this work, we…

Read at arXiv (research papers)

Older stories