Built to outpace AI-supported attackers across your software, firmware, and systems. Finding weaknesses, proving impact, acting when necessary. Before adversaries can.
Argus Systems ingests your repositories to map code and data flow, executing guided vulnerability research across source code and binaries alike. Threat modeling, discovery, and verification aren't built on brittle prompts—they are driven by a robust, multi-agent architecture engineered for scale.
Plug in frontier, open-source, or your own LLMs—Argus Systems scales with them seamlessly. Run our deep LLM intelligence over your system and get verified exploits delivered just hours later.
Real findings from the Argus swarm, disclosed in the open.
Attacker-controlled length fields are used directly as the bcmp comparison length, allowing a zero-length authentication bypass.
An over-read of the label-stack array leaks 4 bytes of adjacent kernel stack memory in the ICMP/MPLS error response.
A low-privilege user can supply an arbitrary userid and receive any user's dashboard rendered as a PDF.
OAuth token values are interpolated into a SQL UPDATE with no escaping; stacked queries enable arbitrary SQL execution.
SSH connection-reuse matching in libcurl does not verify the SSH authentication context.
How the bugs were found, and why they mattered.
Zero-length name and password fields made bcmp return 0 unconditionally, bypassing PAP entirely. A flaw introduced in 1999 that survived 27 years.
Zabbix drops an external OAuth server's response straight into a SQL UPDATE with zero escaping, then dismissed it as a non-issue.
ByteRay runs Argus against your products: find weaknesses, prove impact, patch what matters. Open-source projects are reviewed free of charge.
Book an appointment →