Adversarial Attacks on LLM-As-A-Judge Systems: Insights from Prompt Injections

LLM as judge systems used to assess text quality code correctness and argument strength are vulnerable to prompt injection attacks. We introduce a framework that separates content author attacks from system prompt attacks and evaluate five models Gemma 3.27B Gemma 3.4B Llama 3.2 3B GPT 4 and Clau...

DoomArena: a Framework for Testing AI Agents against Evolving Security Threats

We present DoomArena, a security evaluation framework for AI agents. DoomArena is designed on three principles: 1 It is a plug-in framework and integrates easily into realistic agentic frameworks like BrowserGym for web agents and $τ$-bench for tool calling agents; 2 It is configurable and allows...

Cybersecurity through Entropy Injection: a Paradigm Shift from Reactive Defense to Proactive Uncertainty

Cybersecurity often hinges on unpredictability, with a system's defenses being strongest when sensitive values and behaviors cannot be anticipated by attackers. This paper explores the concept of entropy injection-deliberately infusing randomness into security mechanisms to increase...

Threat actors thrive in chaos

Welcome to this week's edition of the Threat Source newsletter. If there's one thing that threat actors love, it's chaos. Headlines in the news that provoke an emotional response make excellent phishing lures because the intense feelings invoked by a provocative subject line cause our critical...

Capacity is Critical in Riskier Threat Landscape

International cooperation, reporting, and capacity building are critical to enhance cybersecurity defenses. Effective governance in an increasingly risky landscape requires visibility as well as coordinated vulnerability disclosure...

Nearly 24,000 IPs Target PAN-OS GlobalProtect in Coordinated Login Scan Campaign

Cybersecurity researchers are warning of a spike in suspicious login scanning activity targeting Palo Alto Networks PAN-OS GlobalProtect gateways, with nearly 24,000 unique IP addresses attempting to access these portals. "This pattern suggests a coordinated effort to probe network defenses and...

Imperva Protects Against the Apache Camel Vulnerabilities

Introduction: Understanding the Apache Camel Flaw On March 9, 2025, Apache released a security advisory for CVE-2025-27636, a vulnerability in the Apache Camel framework that allows attackers to bypass header filtering via miscased headers. Although rated as moderate, this vulnerability...

Debunking the AI Hype: Inside Real Hacker Tactics

Is AI really reshaping the cyber threat landscape, or is the constant drumbeat of hype drowning out actual, more tangible, real-world dangers? According to Picus Labs' Red Report 2025 which analyzed over one million malware samples, there's been no significant surge, so far, in AI-driven attacks...

AI-Powered Social Engineering: Ancillary Tools and Techniques

Social engineering is advancing fast, at the speed of generative AI. This is offering bad actors multiple new tools and techniques for researching, scoping, and exploiting organizations. In a recent communication, the FBI pointed out: 'As technology continues to evolve, so do cybercriminals'...

SAP Commerce 跨站请求伪造漏洞

SAP Commerce is a suite of cloud-based e-commerce platforms from Germany's SAP. It supports sales management, marketing management, order management and operations management. A cross-site request forgery vulnerability exists in SAP Commerce, which stems from a misconfiguration that can lead to...

AI in Cybersecurity: What's Effective and What's Not – Insights from 200 Experts

Curious about the buzz around AI in cybersecurity? Wonder if it's just a shiny new toy in the tech world or a serious game changer? Let's unpack this together in a not-to-be-missed webinar that goes beyond the hype to explore the real impact of AI on cybersecurity. Join Ravid Circus , a seasoned...

Bot Reporting: 4 Key Differences to Consider

Get a detailed look at Akamai’s bot reporting and the important advantages it offers to organizations that are looking to improve their bot defenses...

Announcing the Adaptive Prompt Injection Challenge (LLMail-Inject)

We are excited to introduce LLMail-Inject, a new challenge focused on evaluating state-of-the-art prompt injection defenses in a realistic simulated LLM-integrated email client. In this challenge, participants assume the role of an attacker who sends an email to a user. The user then queries the...

Hackers Use Corrupted ZIPs and Office Docs to Evade Antivirus and Email Defenses

Cybersecurity researchers have called attention to a novel phishing campaign that leverages corrupted Microsoft Office documents and ZIP archives as a way to bypass email defenses. "The ongoing attack evades antivirus software, prevents uploads to sandboxes, and bypasses Outlook's spam filters,...

U.S. Telecom Giant T-Mobile Detects Network Intrusion Attempts from Wireline Provider

U.S. telecom service provider T-Mobile said it recently detected attempts made by bad actors to infiltrate its systems in recent weeks but noted that no sensitive data was accessed. These intrusion attempts "originated from a wireline provider's network that was connected to ours," Jeff Simon,...

How Is API Abuse Different from Web Application Attacks by Bots?

API abuse and web application bot attacks are often confused. This is understandable, as both involve automated interactions and are usually executed by bots. Both attack vectors are prevalent; criminals are always eager to disrupt the foundations on which businesses base their operations to...

Cyber Story Time: The Boy Who Cried "Secure!"

As a relatively new security category, many security operators and executives I've met have asked us "What are these Automated Security Validation ASV tools?" We've covered that pretty extensively in the past, so today, instead of covering the "What is ASV?" I wanted to address the "Why ASV?"...

mySCADA myPRO Manager

RISK EVALUATION Successful exploitation of this vulnerability could allow a remote attacker to execute arbitrary commands or disclose sensitive information. 2. RECOMMENDED PRACTICES CISA recommends users take defensive measures to minimize the risk of exploitation of these vulnerabilities, such...

Stop LUCR-3 Attacks: Learn Key Identity Security Tactics in This Expert Webinar

Did you know that advanced threat actors can infiltrate the identity systems of major organizations and extract sensitive data within days? It's a chilling reality, becoming more common and concerning by the day. These attackers exploit vulnerabilities in SaaS and cloud environments, using...

Cyber Threats in Costume: When Attacks Hide Behind a Mask

Introduction As Halloween approaches, the idea of costumes and disguises takes center stage, but the spirit of deception isn’t limited to one night. In the digital world, cyberattacks can also wear masks, concealing their true intentions to slip past defenses. Just as a costume can obscure who’s...