S

SafeBreach

Tests defenses with real-world attack simulations

Red Team Developer

Full-TimeUpdated
No salary listed
Mid
Tel Aviv-Yafo, Israel
HybridHybrid role based in Tel Aviv.
CybersecurityOffensive Security & Testing

About the job

Requirements
  • At least 3 years of hands-on experience developing offensive attack components.
  • At least 3 years of Python experience, especially integrating with native modules and supporting research workflows.
  • Experience developing stealthy and evasive attack components, including in-memory execution and user/kernel-mode tooling.
  • Familiarity with endpoint detection and response evasion, API hooking, and direct system call manipulation.
  • Passion for offensive security, deep technical research, and AI coding agents.
  • Ability to operate independently in high-trust environments.
  • Creative and methodical problem-solving with attention to detail.
Responsibilities
  • Lead the design, development, and deployment of production-grade offensive capabilities targeting application and operating system vulnerability exploits.
  • Develop attack components for Windows and Linux, including operating-system-level evasion mechanisms.
  • Implement Python bindings to connect native low-level components with Python-based research tools and automation.
  • Research and develop bypass techniques for modern security controls.
  • Collaborate with the Research Team and other research and development stakeholders to implement and refine offensive concepts.
  • Use AI coding agents to streamline these processes.
Desired Qualifications
  • Deep knowledge of Windows and Linux APIs, security controls, and undocumented behaviors relevant to offensive security.
  • Experience with MITRE ATT&CK and implementing or simulating ATT&CK techniques.
  • Background in red teaming, adversary emulation, or advanced persistent threat simulation.
  • Prior contributions to offensive tooling or internal research and development platforms.
  • Understanding of modern threat actor tactics, techniques, and procedures and practical application in research environments.

About the company

SafeBreach tests client defenses by safely and continuously running real-world attack simulations to quantify a security posture. Its platform integrates with existing security controls like SIEM and SOAR to collect data and provide measurable performance metrics, showing how effective defenses are and where gaps exist. Unlike generic security tests, SafeBreach delivers ongoing, data-driven insight into attack surface and risk. The company works with a wide range of clients—from law firms and healthcare organizations to technology companies—to identify gaps, speed remediation, and reduce overall risk. The goal is to help clients validate and strengthen their security controls, translating findings into actionable improvements and a stronger defensive posture.

Company Size

51-200

Company Stage

Series D

Total Funding

$106.5M

Headquarters

Sunnyvale, California

Founded

2014

Get referred to SafeBreach

See people who can refer or advise you

Benefits

  • Health Insurance
  • Dental Insurance
  • Vision Insurance
  • Unlimited Paid Time Off
  • Flexible Work Hours
  • Remote Work Options
  • Parental Leave
  • Phone/Internet Stipend
  • Professional Development Budget
  • Company Equity
  • Stock Options
  • Employee Referral Bonus

Growth & Insights and Company News

6 month growth

0%

1 year growth

0%

2 year growth

0%
Associated Press
Jul 23rd, 2026
SafeBreach unveils AI agents for exposure management, Anvilogic integration at Black Hat USA 2026

SafeBreach will showcase new AI agent capabilities and an integration with Anvilogic at Black Hat USA 2026 and DEF CON 34 in Las Vegas from 1-9 August. The company is unveiling enhanced functionality for three AI agents within its CTEM Platform: the Analyst Agent, which now integrates with vulnerability management tools; the Validation Agent, extending testing to LLM applications; and the SecOps Agent, which validates detection rules. The new two-way integration with Anvilogic automatically converts SafeBreach attack simulation findings into production-ready detections through AI-powered workflows. SafeBreach Labs researcher Ron Ben Yizhak will present original research across three sessions, covering vulnerabilities in legacy Linux services and privilege escalation in Azure containers.

Associated Press
Apr 22nd, 2026
SafeBreach launches AI-powered continuous threat exposure management solution with Helm agent

SafeBreach has launched an AI-powered Continuous Threat Exposure Management solution designed to help organisations identify, prioritise and remediate cyber risks at scale. The platform operationalises Gartner's five-phased CTEM framework through a unified solution, addressing challenges like AI-generated threats and alert overload. The solution centres on the SafeBreach Helm AI Agent, which integrates with existing security tools including threat intelligence, vulnerability management and external attack surface management systems. It covers all CTEM phases from scoping and discovery through prioritisation, validation and mobilisation, translating findings into actionable guidance. Built on SafeBreach's Exposure Validation Platform, the solution combines breach simulation capabilities with AI-driven remediation. It integrates with workflow management tools including ServiceNow and Jira to enable efficient risk reduction aligned to real-world attacker behaviour.

FYIsoft
Feb 5th, 2026
Infy hackers resume operations with new C2 servers after Iran internet blackout ends

Infy hackers resume operations with new C2 servers after Iran internet blackout ends. The elusive Iranian threat group known as Infy (also known as Prince of Persia) has evolved its tactics as part of an effort to cover its tracks even as it prepares new command and control (C2) infrastructure to coincide with the end of a widespread regime-imposed internet blackout earlier in the month. "Threat actors ceased maintenance on their C2 servers on January 8th for the first time since we began monitoring their activity," said Tomer Bar, SafeBreach's vice president of security research, in a report shared with The Hacker News. "This is the day a country-wide internet shutdown was imposed by Iranian authorities in response to recent protests, suggesting that perhaps even government-affiliated cyber forces had no ability or incentive to carry out malicious activities inside Iran." The cybersecurity firm announced that the hacking group installed a new C2 server and observed new activity on January 26, 2026, the day before the Iranian government eased internet restrictions in the country. This development is significant, especially in that it provides concrete evidence that the adversary is state-backed and backed by Iran. Infy is just one of many state-sponsored hacker groups operating outside Iran, conducting espionage, sabotage, and influence operations in line with Iran's strategic interests. However, it is also one of the oldest and least known groups, having operated quietly under the radar since 2004 through "laser-focused" attacks targeting individuals for information gathering purposes. In a report published in December 2025, SafeBreach revealed new techniques associated with threat actors, including the use of updated versions of Foudre and Tonnerre. The latter appears to be using Telegram bots for issuing commands and collecting data. The latest version of Tonnerre (version 50) is codenamed Tornado. Continuous visibility into the attacker's activity from December 19, 2025 to February 3, 2026 revealed that the attacker introduced Tornado version 51, which uses both HTTP and Telegram in the C2, and took steps to replace all versions of Foudre and Tonnerre's C2 infrastructure. "Two different methods are used to generate C2 domain names: a new DGA algorithm, and then blockchain data deobfuscation to fix the name," Barr said. "This is a unique approach that we assume is being used to provide more flexibility in registering C2 domain names without having to update your Tornado version." There are also indications that Infy was able to weaponize a one-day security flaw in WinRAR (CVE-2025-8088 or CVE-2025-6218) to extract a Tornado payload on compromised hosts. Changing attack vectors is seen as a way to increase the success rate of campaigns. A specially created RAR archive was uploaded to the VirusTotal platform in mid-December 2025, suggesting that both countries may have been targeted. Inside the RAR file is a self-extracting archive (SFX) containing two files. Tornado establishes communication with the C2 server via HTTP, downloads and executes the main backdoor, and collects system information. If Telegram is selected as the C2 method, Tornado uses the bot API to extract system data and receive more commands. It is worth noting that version 50 of the malware used a Telegram group named سرافراز (lit. "sarafraz", meaning proudly) featuring the Telegram bot "@ttestro1bot" and a user with the handle "@ehsan8999100". In the latest version, another user "@Ehsan66442" has been added in place of the latter. "As before, bot members in Telegram groups still do not have permission to read group chat messages," Bar said. "On December 21, the original user @ehsan8999100 was added to a new Telegram channel named Test with 3 subscribers. The purpose of this channel is still unknown, but we believe it is being used for command and control over the victim's machine." SafeBreach announced that it has successfully extracted all messages in private Telegram groups and now has access to all leaked Foudre and Tonaire files since February 16, 2025. This includes 118 files and 14 shared links containing encrypted commands sent to Tonaire by threat actors. Analysis of this data yielded two important findings. Loading a custom variant of StormKitty infostealer Malicious ZIP file that drops ZZ Stealer ZZ Stealer attack chain and previous iterations of ZZ Stealer using a package named "testfiwldsd21233s" designed to leak data through the Telegram bot API Python Package Index (PyPI) "Very strong correlation" between campaigns targeting repositories "Potential correlation" between Infy and Charming Kitten (also known as Educated Manticore) through the use of ZIP and Windows Shortcut (LNK) files and PowerShell loader techniques "ZZ Stealer appears to be a first-stage malware (like Foudre) that first collects environmental data, screenshots, and steals all desktop files," SafeBreach explained. "Furthermore, upon receiving the command "8==3" from the C2 server, it downloads and executes the second stage malware, also named "8==3" by the threat actor."

Android Authority
Aug 6th, 2025
Poisoned calendar invite shows just how easily Gemini can be tricked to hijack your smart home

At the Black Hat security conference this week, a group of researchers from Tel Aviv University, Technion, and SafeBreach showed how they were able to hijack Gemini using what's known as an indirect prompt injection.

01net
Apr 23rd, 2025
SafeBreach to Showcase New Research, Product Innovations, and ServiceNow CTEM Collaboration at RSAC 2025

SafeBreach to showcase new research, product innovations, and ServiceNow CTEM collaboration at RSAC 2025.