Case Studies - Glog.AI

Case Study: AMD (Advanced Micro Devices)

Securing the Future of High-Performance Computing: Glog.AI & AMD ROCm™

The Partner

AMD is a global leader in high-performance computing, graphics, and visualization technologies. Their ROCm™ (Radeon Open Compute) platform is a foundational open-source software stack for GPU computing, AI, and high-performance computing (HPC) workloads.

The Challenge

Open-source ecosystems of this scale present a massive attack surface. Spanning 343 GitHub repositories and nearly one billion lines of code across C, C++, Python, and Linux kernel drivers, securing the ROCm™ stack required automated, context-aware vulnerability detection and actionable remediation without overwhelming developers with false positives or slowing down agile delivery schedules.

The Solution

Software Product Inventory (SPI): All data was funneled into Glog.AI’s centralized dashboard, a “single pane of glass” that provides real-time inventory, security data, and vulnerability analysis.
Comprehensive Visibility: A complete security map of 343 repositories.
Centralized Management: The ability to track vulnerabilities across disparate projects through the Glog.AI SPI dashboard.
Proactive Remediation: Faster identification of potential risks within the ROCm™ stack and Linux kernel drivers with context-aware remediation advice.

Key Highlights

Massive Scale: Scanned 343 repositories and provided comprehensive, actionable remediation advice.
Centralized Management: Established a centralized inventory and management system for all software products.
Seamless Integration: Automated security processes through direct GitHub workflow integration.
Operational Efficiency: Streamlined the vulnerability resolution process without slowing down agile delivery schedules.
Proactive Security: Empowered the team to adopt a “security by design” approach right at the source code level.

Source

Public blog post: 343 Repositories, One Mission: How Glog.AI is Securing AMD’s Open Source Ecosystem

Case Study: Glog.AI Takes #1 on the RealVuln F1 Leaderboard

Glog.AI ranks #1 on the RealVuln F1 leaderboard with a score of 82.7 — the only scanner to exceed 80% on both precision and recall

Snapshot

Benchmark: RealVuln v3.1 (Python stratum)
Date: September 2026
Number of scanners evaluated: 34
Test set: 66 open-source applications, 133,782 lines of code, 1,902 labelled vulnerabilities
Key Glog.AI results:
- F1 score: 82.7 (#1, 4.4 points ahead of the next scanner)
- Precision: 85.1%
- Recall: 80.5%
- The only scanner to clear 80% on both precision and recall simultaneously
- 47% fewer false positives than the next-cleanest scanner at the same detection rate

Challenge

Most code analysis tools (SAST) either generate too many false positives or miss a significant portion of real vulnerabilities. Developers waste time on noise, and AppSec teams lose trust in the results. Traditional rule-based scanners (SonarQube, Semgrep, etc.) rank at the bottom of public leaderboards, while many AI-based approaches sacrifice either precision or coverage.

Solution

Glog.AI Software Security was evaluated on the independent RealVuln benchmark (v3.1). The platform builds a semantic graph of the entire application (data and control flow across files and functions) and uses AI only for final judgment calls (exploitability, deduplication), not for pattern matching. The evaluation consisted of a single complete run across all 66 applications, scored with the benchmark’s official matcher against its published ground truth.

Results

Metric	Glog.AI	Next (Kolega DevSec Max)	Daybreak Blue	Difference
F1 score	82.7	78.3	77.7	+4.4
Precision	85.1%	70.1%	75.0%	+15.0 pp
Recall	80.5%	88.7%	80.5%	—
False positives	269	721	511	−47%
True positives	1,532	1,688	—	—

Additional key findings:

The only scanner out of 34 that exceeds 80% on both precision and recall.
At the same detection level (≈80.5% recall), it produces 47% fewer false positives than the next-cleanest scanner.
Across 74 CWE families: 39 with 100% precision, 24 perfect on both precision and recall.
Performance remains stable on both real (human-authored) and generated code.

Why this matters

F1 score penalizes both excessive noise and missed vulnerabilities. Glog.AI is the only scanner that maintains a high level of both detection and result cleanliness. This directly addresses the biggest pain point in AppSec: developers receive fewer false positives, and AppSec teams gain more confidence in the findings.

Source

Public RealVuln benchmark (Kolega Labs, Apache 2.0). Results and leaderboard are publicly available.
Full details: https://www.glog.ai/glog-ai-takes-1-on-the-realvuln-f1-leaderboard/

Case Study: Glog.AI Helped Easy Aerial Achieve Blue UAS Cleared List Inclusion

Glog.AI helped Easy Aerial achieve Blue UAS Cleared List inclusion through AI-driven software security analysis

Snapshot

Customer: Easy Aerial (US provider of autonomous drone-in-a-box solutions)
Industry: Aerospace / Defense / Unmanned Systems
Outcome: Successful inclusion on the U.S. Department of Defense Blue UAS Cleared List
Platforms: Osprey (Free Flight Version) and Sparrow (Free Flight Version)
Key contribution: AI-driven vulnerability analysis, contextual remediation, and significant reduction of false positives
Date of post: June 2025

Challenge

Easy Aerial needed to meet the rigorous cybersecurity and data privacy standards required for inclusion on the Blue UAS Cleared List (managed by the Defense Innovation Unit / later DCMA). The process demands verified cyber-secure software, NDAA compliance, and successful third-party or DIU assessment. Identifying genuine software vulnerabilities without overwhelming the development team with false positives was critical to passing the verification process efficiently.

Solution

Glog.AI supported Easy Aerial by applying its AI-driven software security platform and expert analysis to their codebase:

Analyzed software for potential weaknesses and vulnerabilities early in the development lifecycle
Performed contextual vulnerability analysis that understood the specific architecture and code of Easy Aerial’s systems
Reduced false positives compared to generic tools, saving the development team significant time
Helped ensure the systems met the high cybersecurity and data privacy requirements of the U.S. DoD

Results

Easy Aerial successfully achieved inclusion on the Blue UAS Cleared List. Both the Osprey and Sparrow platforms are listed. This enables DoD and other federal customers to procure and operate the systems without additional exceptions to policy. Easy Aerial is also ISO 9001 and AS9100 certified, with all products NDAA compliant.

Why this matters

Blue UAS clearance is a high bar for cybersecurity in unmanned systems. This case demonstrates Glog.AI’s ability to support companies in regulated, high-assurance environments (defense, critical infrastructure) by delivering precise, low-noise vulnerability analysis and remediation support that helps teams meet strict compliance requirements.

Source

Public blog post: https://www.glog.ai/glog-ai-helped-easy-aerial-achieve-blue-uas-cleared-list-inclusion/
Easy Aerial platforms are publicly listed on the official Blue UAS Cleared List.

Case Study: GLOG.AI® NetEcho™ — 10 Gbit/s AI Network Anomaly Detection Benchmark

Seeing the Unseen at Wire(less) Speed: How GLOG.AI® NetEcho™ Delivers Passive AI Network Anomaly Detection

Agentless, AI-driven network anomaly detection at up to 10.5 Gbit/s on just 2 vCPU — with zero packet payloads leaving your network.

Snapshot

Product: GLOG.AI® NetEcho™ (part of the unified GLOG.AI Glogosphere platform)
Target Environments: Telecommunications operators, financial institutions, critical infrastructure, and enterprise networks (wired and Wi-Fi / IoT traffic via switch SPAN/mirror ports)
Sensor Footprint Tested: Single lightweight container with 2 vCPU and 2 GiB RAM
Date: October 2026
Key Benchmark Results:
- Wire-Speed Capture: 93.9% to 100% packet capture up to 10.5 Gbit/s (~890,000 packets/sec) using ~0.6 of one CPU core; ~75% capture under 20 Gbit/s traffic bursts
- DDoS & Flood Resilience: Sustained SYN floods reaching 650,000 new flows/sec and bursts of 10,000 flow records/sec with zero dropped or unaccounted records
- 100% Telemetry Integrity: Exact match between sensor logs and NetEcho server across 3,322,311 sessions and 1,727,099 anomalies
- Detection Latency: 7.5 seconds average (10.6 seconds max) from packet arrival to AI classification
- Zero-Impact & Sovereign by Design: Passive out-of-band sensor sends zero packet payloads off-site

Challenge

Enterprises, telecom operators, and critical infrastructure providers need deep east-west and north-south visibility across wired, wireless (Wi-Fi), and unmanaged IoT devices without introducing inline latency, deploying endpoint agents everywhere, or exporting sensitive packet payloads outside their security perimeter. Furthermore, lightweight network sensors often fail, drop telemetry, or run out of memory when subjected to 10+ Gbit/s traffic spikes and volumetric SYN flood attacks.

Solution

GLOG.AI® NetEcho™ uses a three-tier, sovereign-telemetry architecture engineered for speed, resilience, and privacy:

Passive Sensor (Your Network): Connects to a switch mirror (SPAN) or wireless-controller port out-of-band, never sitting in the production traffic path and sending zero packet contents off-site.
Dedicated NetEcho Server (On-Premises or Isolated Cloud): Performs AI-driven behavioral analysis and anomaly classification while keeping sensitive telemetry inside your perimeter.
GLOG.AI® Glogosphere Console (Unified SOC Interface): Correlates network anomalies with Security Predictions and Software Security to calculate a business-weighted Asset Risk Score (0–100) across code, external exposure, and live network traffic.

Verified Benchmark Results

Test Scenario (2 vCPU / 2 GiB Sensor)	Reports	Sessions	Anomalies	Sensor-to-Server Verification
10 Gbit/s load ramp	16	374	73	Exact match
10 Gbit/s load ramp (repeated)	16	381	73	Exact match
SYN flood ramp and soak (up to 650k flows/s)	47	416,184	143,156	Exact match
Classification burst & 4 forced process kills	17	103,291	59,779	Exact match
Flood, then frozen capture	39	364,848	121,755	Exact match
2-hour continuous load soak	241	2,437,233	1,402,263	Exact match (0 restarts, 0 OOM)
Total Verified Suite	376	3,322,311	1,727,099	All exact (0 dropped records)

Why this matters

NetEcho proves that organizations do not need heavy hardware appliances or invasive packet inspection to achieve 10+ Gbit/s AI network anomaly detection. Whether strengthening operational resilience for EU DORA, NIS2, EU CRA, SOC 2, PCI DSS, WLA SCS, and ISO/IEC 27001, or eliminating blind spots across core enterprise and telecom infrastructure, NetEcho delivers verified high-speed detection with complete data sovereignty.

Source

Full benchmark details: https://www.glog.ai/netecho-ai-network-anomaly-detection-10gbps-benchmark/

Case Study: EuroICC

Securing Smart Hotel Control and Embedded Automation Systems for EUROICC

About the Partner

EUROICC is a hardware and software manufacturer, R&D outsourcing partner, and system integrator in automation. Their Smart Hotel Control products enable business efficiency and guest comfort, and their ToCCata guest room management system incorporates room monitoring and hotel management.

Business Challenge

The Guest Room Management System based on the C series of micro-controllers is a flexible solution applicable for different size hotels—from small inns to luxurious hotels. Securing embedded C/C++ firmware and connected management software required identifying real vulnerabilities with high precision and providing context-aware remediation without disrupting product engineering cycles.

Results

Glog.AI analyzed EUROICC’s software and firmware codebase, eliminating false positive noise and providing precise, context-aware remediation advice that strengthened the security posture of their connected automation and hospitality systems.

About Glog.AI

Glog project is focused on research and development of a solution which gives remediation advice for security vulnerabilities in software code based on context. Moreover, it is capable of automatically fixing those vulnerabilities. We are developing such a solution based on machine learning and AI. Agility in software security is becoming a reality! More info at www.glog.ai.

General use cases

Select your scenario to see how we can help:

Software Development Company: Your product is deployed in a production environment. Hackers quickly spot a vulnerable product and exploit vulnerabilities. Data are stolen, altered, or damaged by unauthorized parties—malicious actors. You face liability toward customers and regulators. Your reputation is ruined or damaged. What to do? You need to implement software security practices and processes as soon as possible. We can help. Feel free to contact us. Glog solution helps you to reduce or eliminate security debt. Otherwise, it will grow and can reach a level which you might not be able to resolve in the future.
Startup: Let’s say you are the founder of a startup. With your co-founders and development team you rush to create your MVP – Minimum Viable Product and to put it live. Hackers spot that your MVP is live. For them it is also MVP, but they see it as the Most Vulnerable Product! Regulators see it as an MLP – Massive Liability Product. What to do? You need to implement software security practices and processes as soon as possible. It should be part of the process from the beginning! Contact us to help you.
Software User: You are a company or individual using software from a vendor and want to make sure software is as secure as possible as well as compliant with relevant security and privacy standards and regulations. Ask your software provider or vendor if they have security as part of their SDLC – Software Development Lifecycle. Have they performed necessary security testing and remediated vulnerabilities? What to do if they haven’t done it? Direct them to us to help them secure their software and software supply chain.
Regulatory Body: You are a regulatory body and want to make sure software is secure according to standards and regulations. Ask your software provider or vendor if they have security as part of their SDLC – Software Development Lifecycle. Have they performed necessary security testing and remediated vulnerabilities? Can they prove it? What to do if they haven’t done it? Direct them to us to help them secure their software and software supply chain.
Compliance: With proper software and security processes and practices in place, as part of your SDLC – Software Development Lifecycle, you are much better positioned to be compliant. Contact us to help you.

False Positives Challenge

Leading provider of SAST tools on the market asked us to analyze their scan results and security vulnerabilities findings in two popular projects. Here are results (names and results redacted due to confidentiality and sensitivity). Here are results of Glog solution work on it:

Applications AH and OAM scanned by leading SAST tool on the market and 12,494 security findings reported in total by that tool.
Glog confirmed a total of 49 true positive (TP) findings and offered very precise remediation advice.
Work in progress with a couple of popular tools including messengers and similar.

Glog.AI Products & Services Matrix

This matrix provides a structured overview of Glog.AI’s portfolio, highlighting the AI-centric nature and specific value proposition of each offering.

Product / Service Name	Core Functionality	Key AI/ML Differentiators	Target Use Case
Glog.AI – Software Security	Automated code vulnerability identification and remediation during development	AI-powered contextual analysis, automated fixing, false positive reduction	DevSecOps, Application Security
NetEcho – Network Security	Predictive network and endpoint threat detection, anomaly identification	AI-driven behavioral analysis, threat forecasting	Network Security Monitoring, Endpoint Protection
Security Predictions	Anticipation of emerging attack patterns and potential targets	Advanced AI analysis of diverse threat intelligence feeds, proprietary prediction algorithms	Proactive Threat Intelligence, Strategic Security Planning
vSOC	Virtual Security Operations Center services	Leverages NetEcho & Security Predictions, AI-assisted analysis by human experts	Continuous Security Monitoring, Incident Response
Penetration Testing	AI-enhanced vulnerability identification and assessment	More efficient, comprehensive, and insightful analysis through AI assistance	Security Audits, Vulnerability Management
Trainings	Secure software development education	Practical, up-to-date secure coding practices	Developer Upskilling, Building Security Culture

Articles

Discover how Glog.AI enhances cybersecurity and software security. Our articles delve into real-world examples and explain the ‘how’ behind our solutions.