<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Gultekin Butun on Medium]]></title>
        <description><![CDATA[Stories by Gultekin Butun on Medium]]></description>
        <link>https://medium.com/@gbutun?source=rss-08e3ce050892------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*mOhYWm6It9TaJI6F</url>
            <title>Stories by Gultekin Butun on Medium</title>
            <link>https://medium.com/@gbutun?source=rss-08e3ce050892------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 16:05:44 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@gbutun/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Threat Management in Microsoft Sentinel: A Structured Overview]]></title>
            <link>https://gbutun.medium.com/threat-management-in-microsoft-sentinel-a-structured-overview-9e1b4a21a400?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/9e1b4a21a400</guid>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Mon, 14 Sep 2026 14:24:12 GMT</pubDate>
            <atom:updated>2026-09-14T14:24:12.792Z</atom:updated>
            <content:encoded><![CDATA[<p>It has been a while, now back again with some exiting topics. This time we will go into the world of SIEM or better Sentinel.</p><p>If you open Microsoft Sentinel for the first time, “Threat management” is just another entry in the left navigation pane. Spend a few weeks in the product, though, and you realize it’s not a menu item — it’s an entire discipline, bundled into one umbrella category. Under it sit Incidents, Workbooks, Hunting, Notebooks, Entity Behavior, Threat Intelligence, MITRE ATT&amp;CK (a very important topic as well which I want to cover as well), and SOC Optimization. Most people only ever touch the first one.</p><p>This post walks through what each piece actually does, how they connect, and where the boundaries sit — because those boundaries are exactly where most of the confusion (mine included) tends to happen.</p><p><strong>The Core Distinction: Reactive vs. Proactive</strong></p><p>Everything under Threat Management splits cleanly into two categories:</p><ul><li><strong>Reactive</strong> — something already triggered a detection, and you’re responding to it. This is Incident Management.</li><li><strong>Proactive / contextual</strong> — work that happens independent of whether anything has alerted yet. This is everything else: Hunting, Threat Intelligence, Entity Behavior, MITRE ATT&amp;CK, SOC Optimization.</li></ul><p>The reason this split matters: incident management, by definition, only ever shows you what your rules were already written to catch. If a technique has no rule behind it — or the rule exists but got disabled — the evidence can sit in your raw data indefinitely, and nothing will surface it. Proactive threat management exists specifically to cover that blind spot.</p><p><strong>Incident Management: The Reactive Core</strong></p><p>An <strong>incident</strong> is the actual unit of work in Sentinel. It’s not the same as an alert — an alert is a single detection event; an incident is one or more alerts grouped together based on shared entities (same user, host, or IP address), within a defined time window.</p><p>This grouping happens automatically, by default. Every analytics rule has an “incident creation” toggle that’s on unless you turn it off. When a new alert shares entities with an alert already sitting in an open incident, Sentinel folds it in rather than creating a duplicate. Fusion — Sentinel’s built-in ML correlation engine — does this more aggressively, correlating alerts across entirely different products into a single multistage-attack incident.</p><p>Where does a human fit into this? Correction and judgment, not primary creation:</p><ul><li>Manually merging two incidents Sentinel treated as separate but are actually related</li><li>Manually adding an alert to an existing incident</li><li>Creating an incident from scratch, with no underlying alert (something escalated outside the platform)</li><li>Promoting a hunting bookmark into a trackable incident</li></ul><p>Each incident carries a severity (inherited from the highest-severity alert it contains) and a status (New → Active → Closed), and every closure requires a classification — True Positive, Benign Positive, or False Positive. That classification data isn’t just record-keeping; it’s the feedback signal that tells you which rules are noisy and which are earning their keep.</p><p><strong>What Actually Generates the Alerts</strong></p><p>Analytics rules are the mechanism behind almost every alert:</p><ul><li><strong>Scheduled rules</strong> — your own (or a template’s) KQL, run on an interval you define</li><li><strong>NRT (Near-Real-Time) rules</strong> — a fixed 1-minute cadence, single data source, no joins — built for time-critical, simple-match detections where a 5-minute lag is too slow</li><li><strong>Microsoft Security rules</strong> — a pass-through: every alert a connected Microsoft product (Defender for Endpoint, Identity, Cloud Apps, Office 365, Entra ID Protection, Defender for Cloud) already generated gets imported as-is, no custom logic needed</li><li><strong>Fusion</strong> — a single built-in rule (you can only ever have one active instance) that correlates alerts from Scheduled and Microsoft Security rules into higher-fidelity, multistage incidents.</li></ul><blockquote><strong>Note:</strong> if your workspace is onboarded to the unified Defender portal, classic Fusion may not be available at all — Defender XDR’s own native correlation takes over that role.</blockquote><ul><li><strong>ML Behavior Analytics</strong> — built-in anomaly detections like anomalous SSH login patterns</li><li><strong>Threat Intelligence rules</strong> — match ingested IOCs against your ingested data</li></ul><p>A detail worth internalizing: Content Hub solutions ship dozens of <strong>Scheduled</strong> rule templates per vendor (Cisco, Fortinet, Palo Alto, and so on). None of these are “Fusion rules” — they’re normal Scheduled rules whose <em>output</em>, once properly entity-mapped and MITRE-tagged, becomes fuel Fusion can correlate. One correlation engine, many producers feeding it.</p><p><strong>Content Hub vs. Data Connectors: Two Separate Layers</strong></p><p>This distinction trips up more people than anything else in Sentinel, so it’s worth stating plainly:</p><ul><li><strong>Data Connector</strong> = the pipe. Its only job is moving data into your workspace tables. “Connected” means rows are landing somewhere.</li><li><strong>Content Hub solution</strong> = the packaging. It bundles the connector definition with analytics rules, hunting queries, parsers, and workbooks.</li></ul><p>For newer connectors, installing the solution is what makes the connector appear at all — you install, then go configure the connector separately (a Data Collection Rule for agent-based sources, a consent flow for service-to-service ones). For legacy connectors that predate the Content Hub model (Defender XDR being a common example), the connector can be fully Connected while its corresponding solution still shows “Not installed” in Content Hub. That’s not a bug — it’s just two independent tracking systems that were never retroactively synced. The connector keeps working either way; installing the solution afterward only adds the missing analytics/workbook content, without touching the live connection.</p><p><strong>Entity Behavior Analytics (UEBA): Scoped, Not Universal</strong></p><p><em>UEBA is off by default</em> and, once enabled, only analyzes the specific data sources you explicitly select — not every table in your workspace. Supported sources include Entra ID sign-in and audit logs, Azure Activity, Windows Security Events, AWS CloudTrail, GuardDuty, GCP Audit Logs, Okta, and a handful of specific vendors within CommonSecurityLog (Check Point, Fortinet, Zscaler, CyberArk, Palo Alto). Plain Syslog or an unsupported CEF vendor is invisible to UEBA even with the feature switched on.</p><p>The output lands in its own set of tables (BehaviorAnalytics, IdentityInfo, UserPeerAnalytics) and is additive — it never modifies your source data, and it’s billed as its own ingestion.</p><p><strong>Threat Intelligence: Two Jobs, Not One</strong></p><p>TI indicators (ingested via TAXII, the Upload API, or manual entry) do two distinct things once they’re in your workspace:</p><ol><li><strong>Detection</strong> — TI-based analytics rules actively compare your logs against the indicator list and generate new alerts on a match, functioning as its own detection source.</li><li><strong>Enrichment</strong> — passively adding context (reputation, known threat actor, confidence score) to entities you’re already investigating in an incident.</li></ol><p>For a lab or proof-of-concept environment, this is usually more overhead than it’s worth — sourcing a feed and configuring TAXII rarely produces meaningful matches against synthetic traffic. It earns its place once you’re running against real, internet-facing data.</p><p><strong>Hunting: The Deliberate Gap-Filler</strong></p><p>Hunting is the answer to the question: <em>what if the evidence of an attack is already sitting in my data, but nothing alerted on it?</em> Rules can be disabled, thresholds can be set too high, or a technique might simply have no rule written for it — and from a detection standpoint, all three produce identical silence.</p><p>Hunting is manual and on-demand: you run a KQL query against existing data, right now, looking for a pattern retrospectively. Hundreds of built-in query templates ship with Content Hub solutions (Hunting → Queries), tagged by MITRE technique and filterable by data source. A result worth tracking gets bookmarked; a bookmark worth escalating gets promoted into an incident.</p><p>If a hunt keeps surfacing real positives on a recurring basis, that’s the signal to convert it into a Scheduled rule — turning a manual, occasional check into a permanent, automated one.</p><p><strong>MITRE ATT&amp;CK: Coverage Made Visible</strong></p><p>The MITRE ATT&amp;CK blade renders the full technique matrix and color-codes each cell based on what’s covering it: an active rule, an available-but-unused rule template, a hunting query, or nothing at all. Clicking a covered cell opens the rule’s normal detail/edit page — nothing runs as a side effect of browsing. Clicking a template cell instead offers <strong>Create rule</strong>, which walks through the standard creation wizard pre-filled with the template’s logic and tags.</p><p>The practical value here isn’t the visualization itself — it’s using it to find the grey and blank cells, cross-referencing them against the data sources you actually have connected, and deliberately closing the gaps that matter for your environment instead of leaving detection coverage to chance.</p><p><strong>SOC Optimization: The Feedback Loop</strong></p><p>This is where the whole cycle closes. SOC Optimization surfaces recommendations based on patterns across your own incident history — rules generating excessive false positives, coverage gaps against MITRE techniques, connectors that are ingesting but contributing nothing to detection. It’s the mechanism that turns “we found this manually three times this month” into “we should automate this,” and “this rule fires constantly and gets dismissed every time” into “this rule needs tuning or retirement.”</p><p><strong>Putting It Together</strong></p><p>The full cycle, end to end:</p><p><strong>Connector ingests data → Analytics rule evaluates it → Alert fires → Entity correlation groups it into an incident → Automation rule routes/tags/assigns it → Analyst investigates using the graph, UEBA context, and TI enrichment → Playbook executes the response → Incident closes with a classification → That classification, plus hunting results and MITRE gaps, feeds SOC Optimization → Which informs what gets built or tuned next.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ogAFZ-U9ltPX-BMi-EoU1A.png" /></figure><p>Incident management is the part everyone starts with, because it’s the part that demands attention immediately. But the rest of Threat Management — hunting, TI, UEBA, MITRE coverage, SOC optimization — is what determines whether your incident queue is actually catching what matters, or just catching what you happened to write a rule for six months ago.</p><p>Stay safe 😉</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9e1b4a21a400" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Web Application Firewalls, just some Firewall?]]></title>
            <link>https://gbutun.medium.com/web-application-firewalls-just-some-firewall-30fe062ce9f3?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/30fe062ce9f3</guid>
            <category><![CDATA[web-application-firewall]]></category>
            <category><![CDATA[cloud]]></category>
            <category><![CDATA[waf]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[azure]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Fri, 17 Jul 2026 11:48:36 GMT</pubDate>
            <atom:updated>2026-07-17T11:48:36.976Z</atom:updated>
            <content:encoded><![CDATA[<h3>The WAF Conversation I Have With (almost) Every Client, Eventually</h3><p>Especially if you were in the field from Africa to Europe to Middle East and America, it is always the same conversation; we have a firewall we don’t need a “Web Application Firewall“. After that the frustration starts to explain the customer what firewalls are and why they need it for their application and forth. Mostly I sounded sometimes like a carpet seller in the Grand Bazar, which annoyed me as if I am selling snake oil.</p><p>I’ve lost count of how many Application Gateway and Front Door deployments I’ve walked into where the WAF was sitting there in Prevention mode from day one, blocking legitimate traffic nobody had bothered to baseline first. Somebody ticked a box during provisioning, felt good about it, and moved on. Three weeks later a support ticket lands: “checkout is broken for some users.” Nobody thinks to check the WAF logs until hour four.</p><p>So let’s actually talk about this properly.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1001/1*6zNWd71JpRhG7vclWIos2A.png" /></figure><h3>It’s Not a Firewall, Not Really</h3><p>Calling it a “firewall” undersells what’s happening and oversells what it protects against. A traditional firewall or an NSG looks at IPs, ports, protocols — plumbing. A WAF sits at layer 7 and actually reads the request: the URL, the headers, the cookies, what’s in the body. It’s closer to a very fast, very literal-minded reviewer sitting between your users and your app, checking every request against a list of things that have historically gone wrong.</p><p>That’s the whole trick, and also the whole limitation. It knows what’s gone wrong <em>before</em>. Anything genuinely novel slips past a signature-based engine until someone updates the rule set. This is why I get twitchy when a client treats a WAF deployment as the finish line of an app security conversation rather than one piece of it. Patch your dependencies. Fix your auth. The WAF is not going to save you from a broken session model.</p><p>Let me give you the following analogy; a traditional firewall is layer 3 and 4 where as WAF is layer 7. So if we think it like a deliver company it take the packet to the address written on it safely without knowing what is in it for app, sorry the client. As soon as it arrives; let say a fancy hotel the concierge signs the package, verifies it and delivers it safely to the destination and this concierge is WAF. The point is that, do not expect to deal with packages equally at every level. What makes sense for layer3 and 4 might not be ok for layer 7.</p><h3>The Part Everyone Skips</h3><p>Detection mode first. Always. I don’t care how confident the team is that their app is “clean” — deploy in log-only, let it sit against real production traffic for a week or two, then go through what actually tripped. Nine times out of ten there’s a legacy API endpoint doing something slightly unusual that’s been running fine for two years and would get flatly blocked the second you flip to Prevention.</p><p>The failure mode I see constantly: someone gets a false positive, panics a little, and disables the entire rule <em>category</em> to make the ticket go away. Congratulations, you’ve just reopened the door to every SQLi pattern in that group because one specific parameter on one specific endpoint needed an exception. Scope the exclusion to the parameter. Not the rule group. Not the whole ruleset. The parameter. Below a table that will help in applying the modes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CKQo1AiCxRMBJoPYyVw4Lw.png" /></figure><h3>Azure, and Then Everyone Else</h3><p>If you’re Azure-native, the choice is usually Application Gateway versus Front Door, and it comes down to geography and topology more than features. App Gateway lives inside your VNet — good for internal-facing stuff, regional apps, anything where global edge presence buys you nothing. Front Door sits on Microsoft’s backbone and makes sense the moment you’ve got users spread across continents or you actually need serious DDoS absorption at the edge rather than deeper in your network.</p><p>But here’s the thing I say in almost every proposal now, and it tends to land better with banks and regulated clients than anything else I say in the whole meeting: your WAF choice should follow your traffic, not your cloud provider’s marketing. If you’re genuinely multi-cloud, or you just don’t want your edge security layer married to the same vendor as your compute, Cloudflare is a completely legitimate answer — arguably the more disciplined one. It sits in front of your origin no matter where that origin lives. For a client thinking seriously about exit costs and vendor lock-in, that’s not a minor point, it’s often the deciding one.</p><p>F5 still shows up more than people expect, especially in banking. If a client’s been running anything on-prem for more than a decade, there’s a real chance F5 is already sitting somewhere in their stack, quietly doing this job, and the “new” WAF conversation is actually a hybrid conversation nobody framed correctly. Ask before you assume greenfield. I’ve seen proposals fall apart because nobody asked.</p><h3>What I Actually Tell People</h3><p>Stop thinking of the WAF as a checkbox and start thinking of it as a living piece of config that needs the same change discipline as anything else in production. Baseline it. Tune it narrowly. Route the logs somewhere you’re actually watching — Sentinel, whatever your SIEM is — instead of leaving them to rot in a portal nobody opens.</p><p>And if a vendor’s pitch deck doesn’t mention bot traffic at all, ask why. Credential stuffing and scraping make up a genuinely large share of what a modern WAF needs to catch, and it’s the part that gets left out of the slide when the vendor’s core strength is somewhere else.</p><p>That’s more or less the whole talk. It’s not complicated. It’s just skipped, constantly, because everyone’s in a hurry to tick the box and move to the next line item on the architecture diagram.</p><p>I hope that gives an idea in making decision in security. Stay safe</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=30fe062ce9f3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Tahmine Değil Kanıta Dayanan Çoklu Ajanlı “AI Recon” Aracını Nasıl İnşa Ettim]]></title>
            <link>https://gbutun.medium.com/tahmine-de%C4%9Fil-kan%C4%B1ta-dayanan-%C3%A7oklu-ajanl%C4%B1-ai-recon-arac%C4%B1n%C4%B1-nas%C4%B1l-i%CC%87n%C5%9Fa-ettim-2e834e86677b?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/2e834e86677b</guid>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[pentesting]]></category>
            <category><![CDATA[vulnerability]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Tue, 14 Jul 2026 09:28:24 GMT</pubDate>
            <atom:updated>2026-07-14T09:28:24.521Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SGqqs7JZf77Q7PDZvWi-0Q.png" /></figure><p>Biraz ara verdikten sonra yine heyecan verici bir konuyla karşınızdayız. Bu kez konumuz, yapay zekâyı siber güvenlikte nasıl verimli kullanabileceğimiz. Merak etmeyin, bu yazı süslü bir yapay zekâ blog yazısı değil.</p><p>Benim gibi siz de muhtemelen yapay zekânın tüm pentest sürecini tek başına yaptığı iddia edilen paylaşımları gördünüz. Bazıları fazla iyi, hatta “fazla fantastik”. Bu yazıda size <strong>gerçekçi</strong> bir çözüm göstereceğim.</p><h3>Her Pentest Aynı Başlar: Nmap. Bekle. Nikto. Bekle. WhatWeb, Gobuster, FFUF, WPScan, SSLScan…</h3><p>Yedi araç, yedi terminal çıktısı… Sonra bir saat boyunca hepsini okuyup risk tablosu ve aksiyon notlarıyla bir rapora dönüştürmek. Hep aynı, tekrar tekrar. Otomasyon çağında bu süreci kolaylaştırmak için elimizdeki imkânları kullanmak mantıklı.</p><p>Zor değil. Sadece mekanik. Ben de otomatikleştirdim.</p><p>Ama bunu “ağınızın muhtemelen şöyle olduğunu tahmin eden” bir sohbet botuyla değil; <strong>araçları gerçekten çalıştıran ve yalnızca ürettikleri çıktılar üzerinde akıl yürüten</strong> bir pipeline ile yaptım.</p><h3>Ne Çözüyor ve Size Ne Sağlıyor?</h3><p>Sorun “tarama araç gerektiriyor” değil. Sorun, tarama araç gerektiriyor ve sonra birinin yedi farklı çıktı formatını okuyup bunları tutarlı bir rapora dönüştürmesi. Bu kısım, araçlar ne kadar hızlı olursa olsun sıkışmıyor.</p><p>Bu çözüm size şunları sağlıyor:</p><ul><li>Angarya işlerin bitmiş hâli. Tek komutla planlama, gerçek tarama ve bitmiş rapor.</li><li>Kanıta dayalı rapor. Her bulgu, bir komutun ürettiği gerçek bir satıra dayanıyor. “Model öyle tahmin etti” yok.</li><li>Belirsizlikte dürüstlük. CVE’ler “Confirmed”, “Candidate” veya “Needs-lookup” olarak işaretleniyor.</li><li>Güvenlik bariyeri. Araçlar sadece sıkı bir allowlist içinden çalışıyor ve yalnızca izin verdiğiniz hedeflere yönlendiriliyor.</li><li>Çalışan bir mimari. Gerçek komut çıktısını LLM’e verip yapılandırılmış veri almak isteyen herkes için yeniden kullanılabilir bir yaklaşım.</li></ul><h3>Nasıl Çalışıyor? (Detay olmadan)</h3><p>Ben de çok derine dalınca sıkılanlardanım, o yüzden basit tutalım. Bu çözümü geliştirirken bir YouTube videosundan ilham aldım ama oradaki yaklaşımın bazı eksikleri vardı.</p><p>Sistem üç aşamadan oluşuyor:</p><h3>1. Planlama</h3><p>Hedefe göre bir tarama planı oluşturuyor: hangi araçlar, hangi sırayla. Model komutların tam sentaksına karar vermiyor; sabit bir komut şablonu gerçek hedefe uygulanıyor.</p><h3>2. Çalıştırma (Güvenlik Kapısıyla)</h3><p>Her komut çalışmadan önce iki kurala göre kontrol ediliyor:</p><ul><li>Binary allowlist’te mi? (nmap, nikto, whatweb, gobuster, ffuf, wpscan, sslscan)</li><li>Hedef özel/lab/izinli mi? Herhangi biri geçmezse tarama başlamadan duruyor.</li></ul><h3>3. Analiz ve Raporlama</h3><p>Her aracın gerçek stdout/stderr çıktısı modele veriliyor. Önceden özetlenmiş bilgilerle birlikte (portlar, versiyonlar, TLS bulguları, keşfedilen yollar). Model yapılandırılmış bulgular döndürüyor, ikinci aşama ise bunu rapora dönüştürüyor: özet, risk tablosu, detaylı bulgular, aksiyon planı.</p><p>Hepsi bu. Plan ağınıza dokunmuyor. Çalıştırma kilitli. Rapor sadece olanı anlatıyor.</p><h3>İki Farklı Yapay Zekâ Sistemi Kullanılıyor</h3><p>Bunu açıkça söylemekte fayda var: iki farklı yapay zekâ görev yapıyor.</p><ul><li><strong>Claude Code</strong>: Aracı geliştirirken mimariyi, guardrail’leri ve testleri yazmama yardımcı oldu.</li><li><strong>GPT-4o</strong>: Aracı çalıştırdığınızda tarama çıktısını okuyup raporu oluşturan model.</li></ul><p>Kod yazarken biri, çalışırken diğeri. Konsept her modelle uygulanabilir.</p><h3>Nasıl Kurulur?</h3><p>Çözüm burada: <a href="https://github.com/gbutun/multi_agent_ai_cyber_code?utm_source=copilot.com"><strong>https://github.com/gbutun/multi_agent_ai_cyber_code</strong></a></p><p>Kurulum:</p><p>bash</p><pre>pip install ankacore-attack-chain<br>export OPENAI_API_KEY=&quot;sk-...&quot;<br>attack-chain --target dvwa.local</pre><p>Hepsi bu. Planı oluşturur, komutları doğrular, araçları çalıştırır ve dört dosya üretir: plan, ham çıktılar, AI analizi, final rapor.</p><p>Araçlar yoksa setup_req.sh ile kurabilirsiniz.</p><h3>İki Faydalı Parametre</h3><ul><li>--dry-run: Ağa dokunmadan planı ve doğrulamayı yürütür.</li><li>--report-only: Mevcut tarama verisini kullanarak sadece AI analizini yeniden çalıştırır.</li></ul><h3>Gereksinimler</h3><ul><li>Python 3.9+</li><li>OpenAI API anahtarınız</li><li>Yedi araç (nmap, nikto, whatweb, gobuster, ffuf, wpscan, sslscan)</li><li>Yetkili bir hedef</li></ul><h3>Bir Sorun ve Öğrettiği Şey</h3><p>ffuf başarıyla çalışsa bile ASCII banner’ını stderr’e basıyor. İlk sürümde stderr doluysa onu özetlemeye çalışıyordum. Banner’daki ters eğik çizgiler JSON’a yanlış kaçtı, model yapılandırılmış veri yerine düz metne döndü — sessiz bir hata.</p><p>Ders: LLM’e ham araç çıktısı verip yapılandırılmış veri bekliyorsanız, <strong>parsing’i test edin</strong>. Hata sessiz olur, sonuç boş döner.</p><h3>Sonuç</h3><p>Özetle, klasik tarama araçlarını çalıştıran bir plan üretiyoruz ve çıktıları yapay zekâya analiz ettiriyoruz. Parametre üretmek sihir değil; sihir, <strong>çıktıların doğru zamanda ve doğru şekilde AI tarafından değerlendirilmesi</strong>. Sonuçta riskler, zafiyetler ve CVE’lerle dolu bir rapor elde ediyoruz. Mükemmel değil ama birçok şeyi çözüyor.</p><p>Umarım faydalı bulursunuz. Güvende kalın.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2e834e86677b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How I built a multi-agent AI recon tool that refuses to guess]]></title>
            <link>https://gbutun.medium.com/how-i-built-a-multi-agent-ai-recon-tool-that-refuses-to-guess-04caf740ae7b?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/04caf740ae7b</guid>
            <category><![CDATA[vulnerability]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[pentesting]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Tue, 14 Jul 2026 09:05:36 GMT</pubDate>
            <atom:updated>2026-07-14T09:05:36.665Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SGqqs7JZf77Q7PDZvWi-0Q.png" /></figure><p>After some break we are back again with some exiting topics. This time the topic is to how to use AI efficient in cybersecurity. Do not worry this is not one of the blog posts written with a sugar coating AI language :) .</p><p>Like me you have seen many post where pen testing is supposedly all done by AI etc. Some of the are too good to be true and are “too fantastic”. In this blog post I will share with you a realistic solution.</p><p>Every lab pen-test starts the same way. Run nmap. Wait. Run nikto. Wait. Run whatweb, gobuster, ffuf, wpscan, sslscan. Seven tools, seven blocks of terminal output — then an hour of reading through all of it and writing it up into something with a risk table and remediation notes. It is all the same again and again and especially in the age of automation it only makes sense to use what we are offered.</p><p>None of that is hard. It’s just mechanical. So I automated it.</p><p>Not with a chatbot that guesses what your network probably looks like. With a pipeline that actually runs the tools, and only lets AI reason over what they actually printed.</p><h3>What it solves, and what you actually get out of it</h3><p>The problem isn’t “recon takes tools.” It’s that recon takes tools <em>and then</em> a person has to read seven different output formats, cross-reference them, and turn that into a coherent writeup — and that part doesn’t compress, no matter how fast the scans run.</p><p>What this gets you:</p><ul><li>The grunt work done for you. One command plans the recon, runs it for real, and hands you a finished report — not a to-do list.</li><li>A report grounded in evidence, not vibes. Every claim traces back to an actual line of tool output. If you ask “where did this finding come from,” the answer is always “this exact command printed this exact thing,” never “the model thought it was likely.”</li><li>Honesty about uncertainty. CVEs get labeled Confirmed, Candidate, or Needs-lookup depending on how sure the evidence actually makes it — instead of either overclaiming or silently dropping anything it’s not 100% sure about.</li><li>A safety net that doesn’t rely on you remembering it. It physically cannot run a tool outside a small allowlist, or point that tool at a target you didn’t authorize. That’s enforced in code, not just written in a warning banner.</li><li>A pattern you can steal. If you’re building anything that hands real command output to an LLM and expects structured data back, the guardrail and evidence-parsing approach here is the actual reusable idea — more on that below.</li></ul><h3>How it works, without the deep dive</h3><p>I am also one of the type getting bored when diving in too much so let us keep it simple. Not forget to mention that I was inspired by a You Tube video in developing this solution but the You Tube one has some flaws.</p><p>Three stages, each with one job:</p><ol><li><strong>Plan</strong>. Given a target, it produces a recon plan — which tools, what order. But the tool doesn’t trust the model’s exact command syntax. It swaps in a fixed, hand-written template for the real target, so the model decides <em>that</em> recon happens, not the exact shell command that touches the network.</li><li><strong>Run</strong>, with a gate in front. Before any command executes, it’s checked against two rules: is the binary on a small fixed allowlist (nmap, nikto, whatweb, gobuster, ffuf, wpscan, sslscan — nothing else, ever), and is the target private/lab/explicitly authorized. Fail either check and the whole run stops before anything touches the network.</li><li><strong>Analyze</strong>, then report. The real stdout/stderr from every tool gets handed to the model along with a pre-parsed summary (open ports, versions, TLS findings, discovered paths). It comes back as structured findings with confidence-labeled CVEs, then a second pass turns that into the actual report: executive summary, risk table, detailed findings, remediation roadmap.</li></ol><p>That’s it conceptually. The plan doesn’t touch the network. The execution is locked down. The write-up only ever describes what actually happened.</p><p>One thing worth being upfront about: two different AI systems are involved, doing two different jobs. I built the tool itself with Claude (Anthropic’s Claude Code) as my pair programmer — the architecture, the guardrails, the tests. What the <em>tool</em> calls at runtime, reading your scan output and writing your report, is OpenAI’s GPT-4o. One helped me write the code. The other runs when you actually use it. You could make it work in any way, it is all about the concept.</p><h3>How to deploy it</h3><p>So where is this solution? Check out <a href="https://github.com/gbutun/multi_agent_ai_cyber_code">https://github.com/gbutun/multi_agent_ai_cyber_code</a></p><pre>pip install ankacore-attack-chain<br>export OPENAI_API_KEY=&quot;sk-...&quot;<br>attack-chain --target dvwa.local</pre><p>That’s the whole thing. It generates or reuses a recon plan, validates every command against the safety rules, runs the tools, and writes four files: the plan, the raw scan output, the AI’s structured analysis, and the final report — plain files, nothing hidden behind a dashboard.</p><p>If you don’t already have the seven underlying tools installed, there’s a setup_req.sh you can run first — it checks for each one and installs whatever&#39;s missing through your system&#39;s package manager (apt-get, brew, dnf, or pacman, whichever it finds).</p><p>Two flags worth knowing before your first real run:</p><ul><li>--dry-run walks the entire plan-and-validate pipeline without touching the network. Use this to sanity-check a target before you actually scan it.</li><li>--report-only reuses scan data you already have and just re-runs the AI side. Useful if you&#39;re iterating and don&#39;t want to re-scan the same target every time.</li></ul><h3>Requirements</h3><ul><li>Python 3.9+</li><li>Your own OpenAI API key — the agents call the OpenAI API directly; nothing is proxied or metered for you.</li><li>The seven tools themselves, on your PATH — nmap, nikto, whatweb, gobuster, ffuf, wpscan, sslscan. Standard packages on Kali/Parrot, available via apt/brew elsewhere. Run setup_req.sh and it&#39;ll install whichever ones you&#39;re missing; the chain itself just checks that a command&#39;s binary is allowlisted, it doesn&#39;t install anything mid-run.</li><li>A target you’re actually authorized to test. The guardrails block everything else, but that’s a backstop, not a substitute for having permission in the first place.</li></ul><h3>One thing that broke, worth knowing</h3><p>ffuf prints its ASCII-art banner to stderr even on success. My first pass at summarizing tool output defaulted to reading stderr when it wasn&#39;t empty — reasonable assumption, wrong for this one tool. That banner&#39;s backslashes ended up in what I handed the model, the model copied them into JSON without escaping properly, and the entire structured response silently fell back to raw text — every finding gone, no error, just quietly empty.</p><p>The lesson: if your pipeline hands raw tool output to an LLM and expects structured data back, test the parsing before you trust it. The failure mode is silent. You get a technically successful response missing everything you actually needed.</p><p>That’s the real shape of this project — not “AI writes your pentest report,” but a real tool-execution pipeline with a hard safety allowlist, wrapped in just enough structure that an LLM can turn messy terminal output into something you can act on, and tells you when it isn’t sure instead of making something up.</p><h3>Conclusion</h3><p>Let us wrap this up, we just generate a plan which is running the usual suspects and have it evaluated by AI. The generation of the parameters and feeding the tool is not the magic , the magic is when to have those outputs evaluated by AI and in our case it is Claude. At the end we have a report with risks, vulnerabilities and CVEs. It is not perfect but soves a lot of things.</p><p>I hope you will find it useful, stay safe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=04caf740ae7b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to connect to Azure VMs like with Azure Bastion but cheaper ?]]></title>
            <link>https://gbutun.medium.com/how-to-connect-to-azure-vms-like-with-azure-bastion-but-cheaper-d91aedbfaf4e?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/d91aedbfaf4e</guid>
            <category><![CDATA[connection]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[secure-connectivity]]></category>
            <category><![CDATA[bastion]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Fri, 15 May 2026 11:26:08 GMT</pubDate>
            <atom:updated>2026-05-15T11:26:08.679Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v6QKt-jweeKgZK-qBvZg0Q.png" /></figure><p>Back again, welcome to our blog post on how we can use a cheaper alternative and also secure in Azure to reach Azure VMs with RDP or SSH.</p><p>Through the years working with customers it is always a debate on how to reach Azure VM securely.</p><p>When we say securely we mean:</p><ul><li>VMs have no public ip</li><li>Not using the RDP protocol directly but through any browser</li></ul><p>In fact Azure Bastion is doing that but it has some issues. In one of the latest announcements it was said there is support for integrating with Entra but unfortunately it is not what I expected and came with a lot of issues where we could not even solve it with MS support.</p><p>It is a little pricy. Yes you heard it right, when I look for what we get in return it is not worth it. At the end in my projects I used other alternatives. There are many ways to connect VM fast and secure but today I will give you the most efficient way I have encountered. You may have another approach but this way is just fine for me.</p><p>In the past in my home lab I used Apache Guacamole which is a solution that helps you to connect to various VM source in one place though your browser. So I was thinking , why not have the same solution in Azure.</p><p>I used as IaC Terraform and Guacamole in docker and kept it as simple as is.</p><h3>Architecture Overview</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LTY3HcVdZzPKkudQ8G5vWw.png" /></figure><p>The environment is deployed into a dedicated Azure resource group and virtual network. A Linux VM acts as the public entry point and runs Apache Guacamole using Docker Compose. The Guacamole stack includes the Guacamole web application, guacd, and a MySQL database for Guacamole configuration and connection data.</p><p>A separate Windows Server VM is deployed as the private RDP target. The Windows VM has only a private network interface and no public IP address.</p><p>Network access is controlled with a network security group. Public inbound access is limited to trusted source addresses for Guacamole and administrative SSH. No public RDP rule is configured.</p><h3>Connection Flow</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K_Ta5czHSBqEQ3pjqrKl1A.png" /></figure><h3>Security Benefits</h3><p>This pattern keeps the Windows VM off the public internet while still allowing remote desktop access. RDP is reachable only inside the private network, and users interact with it through Guacamole instead of connecting directly.</p><p>The solution also keeps the Guacamole deployment simple by using the default upstream Guacamole image and theme, avoiding custom branding or patched container images.</p><h3>Final State</h3><p>The final deployment includes:</p><ul><li>A Linux bastion VM running Apache Guacamole</li><li>A private Windows Server VM used as the RDP target (you can ignore that, this is just for testing)</li><li>Docker Compose for Guacamole, guacd, and MySQL</li><li>A public endpoint only for the Guacamole web UI</li><li>No public RDP exposure</li><li>Default upstream Guacamole image and theme</li><li>Terraform-managed infrastructure with targeted deployment support</li></ul><p>This gives a lightweight bastion pattern that is easy to operate, avoids unnecessary customization, and provides controlled browser-based access to private servers.</p><p>Now this is the least minimum, we also should add:</p><ul><li>A firewall before everything with a DNAT rule to the Linux Bastion</li><li>For Guacamole your own theme which should be easy no matter on Docker or not.</li></ul><p>Ok, so where is this solution? Check <a href="https://github.com/gbutun/ankacore-bastion">https://github.com/gbutun/ankacore-bastion</a> , but let say you are busy just reach us out under <a href="http://www.ankacoresystems.com/">www.ankacoresystems.com</a> so we do it for you 😉.</p><p>Stay safe until next time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d91aedbfaf4e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Azure public ip setter]]></title>
            <link>https://gbutun.medium.com/azure-public-ip-setter-68114fb16815?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/68114fb16815</guid>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Tue, 05 May 2026 07:53:16 GMT</pubDate>
            <atom:updated>2026-05-05T07:53:16.621Z</atom:updated>
            <content:encoded><![CDATA[<h3>🔐 A Small Detail, A Big Risk in Azure Firewall IP Management</h3><p>Hello to everyone following AnkaCore,</p><p>We are excited and proud to publish our first blog post. As with our previous content before AnkaCore, our goal here is not to focus on broad or surface-level topics, but rather on more targeted subjects — practical insights that make a real difference and are often overlooked.</p><p>In this post, we focus on exactly that:<br>a detail that is frequently ignored, yet critically important from both a security and operational perspective.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1FdLF4xkOnaAH0Ctu9jncg.png" /></figure><h3>🔍 Problem: A Simple-Looking Approach</h3><p>If you are using Azure and care about security, you are likely familiar with the <strong>built-in firewall (network access control)</strong> mechanisms available in services such as Azure Storage, Key Vault, and others.</p><p>These controls allow you to restrict access to a resource by permitting only specific IP addresses. In most cases, we simply add our own public IP address and block all others.</p><p>At first glance, this approach is simple and effective.</p><p>And it truly is…<br>at least in the beginning.</p><h3>⚠️ Anti-Pattern: IP Sprawl</h3><p>Everything works smoothly — especially when operating from within a corporate network.</p><p>However, the situation changes when you start working remotely.</p><p>There is no guarantee that your public IP address will remain the same (unless you are using a static IP service). One day you have seamless access to your resource, and the next day you suddenly don’t.</p><p>At that point, the typical workflow looks like this:</p><ul><li>Open the Azure portal</li><li>Navigate to the resource (e.g., Azure Storage) and its <strong>Networking / Firewall</strong> settings</li><li>Add the new IP address <em>(often detected automatically and added with a single click)</em></li></ul><p>Problem solved.</p><p>But in reality, it is not solved — it is postponed.</p><p><strong>Over time:</strong></p><ul><li>The IP list keeps growing</li><li>Ownership becomes unclear (who added which IP?)</li><li>Cleanup is rarely performed</li><li>Control is gradually lost</li></ul><p><strong>And most importantly:</strong></p><p>Every newly added IP expands the potential access surface.</p><p><strong>This leads to:</strong></p><ul><li>Increased attack surface</li><li>Higher security risks</li><li>More complex auditing and governance</li></ul><p>What appears to be a practical short-term solution ultimately turns into a long-term security problem.</p><h3>💡 Solution: Azure IP Setter (Automated IP Reconciliation)</h3><p>You don’t need complex systems to solve this problem.</p><p>A simple — but well-designed — automation is enough.</p><p>The <strong>Azure IP Setter</strong> approach does exactly that:</p><ul><li>Retrieves your current public IP address from the runtime environment <em>(e.g., via an external service)</em></li><li>Checks the firewall rules of the target Azure resource</li><li>Updates the IP address if required</li><li><em>(Optional)</em> Cleans up outdated or unnecessary IP entries</li></ul><p>And most importantly, it makes this entire process fully automated.</p><p>Regarding this problem I generated some simple scripts.</p><p>👉 <a href="https://github.com/gbutun/scripts">https://github.com/gbutun/scripts</a><br>contains a total of <strong>four scripts</strong>:</p><ul><li>Two written in <strong>PowerShell</strong></li><li>Two written in <strong>Bash</strong></li></ul><p>While both options are available, I generally prefer <strong>Bash-based solutions</strong>, especially in Linux environments, due to their lightweight and flexible execution model.</p><p>Although the core functionality is provided by setAzureIP.sh | .ps1, I strongly recommend also using setAzureSPN.sh | .ps1.</p><h3>🔐 Authorization and Zero Trust Approach</h3><p>If you use the setAzureIP script alone, it will rely on whichever account you are currently logged in with. If that account does not have the required permissions on the target resource, the operation will fail.</p><p>More importantly, from a <strong>Zero Trust</strong> perspective, we should avoid using highly privileged user accounts for routine operations.</p><p>This is where the setAzureSPN script comes into play:</p><ul><li>It creates a <strong>Service Principal (SPN)</strong> named spn-automation</li><li>Assigns it the <strong>Contributor</strong> role at the subscription level</li></ul><blockquote>(This example focuses on Azure Storage; however, services such as Key Vault, Redis, or others may require different role assignments depending on their access model.)</blockquote><p>💡 <strong>Note:</strong><br>The <strong>Contributor role is sufficient</strong> for managing firewall/network rules.<br>Additional data-plane roles (such as <em>Storage Blob Data Contributor</em>) are not required unless explicitly needed.</p><h3>⚠️ Important Note</h3><p>Since the script performs role assignments, it must be executed using an account with <strong>Owner</strong> privileges.</p><p>However:</p><ul><li>This is a <strong>one-time administrative operation</strong></li><li>After that, all operations can be performed using the SPN</li><li>No further use of highly privileged user accounts is required</li></ul><h3>🔐 Credential Management</h3><p>Credentials such as <strong>Tenant ID, Client ID, Secret, and Subscription ID</strong> are stored locally as environment variables.</p><p>This means:</p><ul><li>They do not leave your execution environment</li><li>However, proper handling and secure configuration practices must still be followed</li></ul><h3>⚙️ How It Works</h3><p>This approach typically consists of the following components:</p><ul><li>A simple <strong>Bash or Python script</strong><br><em>(It can also be implemented in PowerShell; however, Bash provides a more lightweight and flexible model, especially in Linux environments. For Windows users, WSL is a strong alternative.)</em></li><li><strong>Azure CLI (</strong><strong>az) integration</strong></li><li>A <strong>scheduler</strong><br><em>(cron, Azure Automation, or GitHub Actions)</em></li></ul><p>The script runs at regular intervals, ensuring that access is always aligned with your current public IP.</p><h3>🔧 Setup and Scheduling</h3><p>Assuming you have cloned the repository:</p><pre>git clone https://github.com/gbutun/scripts.git</pre><h3>1️⃣ Make the script executable</h3><pre>chmod +x /path/to/setAzureIP.sh</pre><h3>2️⃣ Open your crontab</h3><pre>crontab -e</pre><h3>3️⃣ Add the following line (runs daily at 01:00)</h3><pre>0 1 * * * /usr/bin/bash /path/to/setAzureIP.sh &gt;&gt; /var/log/setAzureIP.log 2&gt;&amp;1</pre><p>This ensures the script runs automatically, updates your IP, and logs execution details.</p><h3>🎯 Benefits</h3><p>The impact of this small automation is bigger than it seems:</p><p>✔ Eliminates IP sprawl<br>✔ Reduces attack surface<br>✔ Removes manual intervention<br>✔ Enables a deterministic access model<br>✔ Improves auditability</p><h3>🧭 Strategic Perspective</h3><p>This approach goes beyond operational convenience.</p><p>It directly contributes to:</p><ul><li>Alignment with <strong>Zero Trust principles</strong></li><li>Enforcement of <strong>least-privilege access</strong></li><li>Improved <strong>cloud governance maturity</strong></li><li>Enhanced <strong>operational excellence</strong></li></ul><h3>🧩 Conclusion</h3><p>Sometimes, the most critical problems hide within the simplest details.</p><p>Dynamic IP management is one of them.</p><p>Left unmanaged, it becomes a growing risk.<br>Handled correctly, it becomes a controlled and automated process.</p><p><strong>A simple script can make a significant difference.</strong></p><p><strong>Stay safe and follow us.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7kuuwB487n4IffFrOcXbig.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=68114fb16815" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Azure public ip setter]]></title>
            <link>https://gbutun.medium.com/azure-public-ip-setter-dd5c57512999?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/dd5c57512999</guid>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Tue, 05 May 2026 07:51:13 GMT</pubDate>
            <atom:updated>2026-05-05T07:51:13.939Z</atom:updated>
            <content:encoded><![CDATA[<h3>🔐 Azure kaynak firewall IP yönetiminde küçük bir detay, büyük bir risk</h3><p>AnkaCore olarak bizi takip eden herkese merhaba.</p><p>İlk blog yazımızı yayınlamanın heyecanını ve gururunu yaşıyoruz. AnkaCore öncesindeki paylaşımlarımızda olduğu gibi, burada da genel ve yüzeysel konular yerine; daha odaklı, pratikte fark yaratan ve çoğu zaman gözden kaçan detayları ele almayı hedefliyoruz.</p><p>Bu yazıda da tam olarak böyle bir konunun üzerine gidiyoruz:<br>Genellikle atlanan, ancak güvenlik ve operasyon açısından kritik öneme sahip bir detay.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1OGri_SG_0y1EUmoZGnZPQ.png" /></figure><h3>🔍 Problem: Basit Görünen Bir Yaklaşım</h3><p>Azure kullanıyor ve güvenliğe önem veriyorsanız, Azure Storage, Key Vault ve benzeri servislerde yer alan <strong>built-in firewall (network access control)</strong> mekanizmasına aşinasınızdır.</p><p>Bu yapı sayesinde, ilgili kaynağa erişimi yalnızca belirli IP adresleri ile sınırlandırabiliriz. Çoğu zaman kendi public IP adresimizi ekler, diğer tüm erişimleri engelleriz.</p><p>İlk bakışta oldukça basit ve etkili bir yaklaşım.</p><p>Ve gerçekten de öyle…<br>en azından başlangıçta.</p><h3>⚠️ Anti-Pattern: IP Sprawl</h3><p>Her şey yolunda gider — özellikle şirket ağı üzerinden çalışıyorsanız.</p><p>Ancak uzaktan çalışmaya geçtiğinizde durum değişir.</p><p>Public IP adresinizin sabit kalacağının bir garantisi yoktur (statik IP hizmeti kullanmadığınız sürece). Bir gün sorunsuz eriştiğiniz kaynağa, ertesi gün erişemediğinizi fark edersiniz.</p><p><strong>Bu noktada tipik senaryo şudur:</strong></p><ul><li>Azure portal açılır</li><li>İlgili kaynağın (örneğin Azure Storage) <strong>Networking / Firewall</strong> ayarlarına gidilir</li><li>Yeni IP eklenir <em>(çoğu zaman IP otomatik algılanır ve tek tıkla eklenebilir)</em></li></ul><p>Sorun çözülür.</p><p>Ama aslında çözülmez — ertelenir.</p><p><strong>Zamanla:</strong></p><ul><li>IP listesi büyür</li><li>Kimin hangi IP’yi eklediği belirsizleşir</li><li>Temizlik yapılmaz</li><li>Kontrol kaybolur</li></ul><p><strong>Ve en önemlisi:</strong></p><p>Her eklenen IP, potansiyel erişim yüzeyini genişletir.</p><p>Bu durum:</p><ul><li>Saldırı yüzeyini artırır</li><li>Güvenlik risklerini yükseltir</li><li>Denetim ve yönetimi zorlaştırır</li></ul><p>Kısa vadede pratik görünen bu yaklaşım, uzun vadede kontrolsüz bir güvenlik problemine dönüşür.</p><h3>💡 Çözüm: Azure IP Setter (Automated IP Reconciliation)</h3><p>Bu problemi çözmek için karmaşık sistemlere ihtiyacınız yok.</p><p>Basit ama doğru bir otomasyon yeterlidir.</p><p><strong>Azure IP Setter</strong> yaklaşımı tam olarak bunu yapar:</p><ul><li>Çalıştığınız ortamdan mevcut public IP adresini alır <em>(örneğin harici bir servis üzerinden)</em></li><li>Azure üzerindeki ilgili kaynağın firewall kurallarını kontrol eder</li><li>Gerekliyse IP adresini günceller</li><li><em>(Opsiyonel)</em> eski ve gereksiz IP’leri temizler</li></ul><p>Ve bu süreci tamamen otomatik hale getirir.</p><p>Bu sorun ile ilgili basit ama etkili scriptler olusturdum.</p><p>👉 <a href="https://github.com/gbutun/scripts">https://github.com/gbutun/scripts</a><br>altında toplam <strong>4 adet script</strong> bulunmaktadır:</p><ul><li>2 adet <strong>PowerShell</strong></li><li>2 adet <strong>Bash</strong></li></ul><p>Her iki seçenek de kullanılabilir olmakla birlikte, ben genellikle <strong>Linux ortamlarında Bash tabanlı çözümleri</strong> tercih ediyorum. Bash, hafif yapısı ve esnekliği ile otomasyon senaryolarında oldukça etkilidir.</p><p>Çözümü doğrudan gerçekleştiren script setAzureIP.sh | .ps1 olmakla birlikte, benim önerim setAzureSPN.sh | .ps1 scriptinin de kullanılmasıdır.</p><h3>🔐 Yetkilendirme ve Zero Trust Yaklaşımı</h3><p>Eğer setAzureIP scriptini tek başına kullanırsanız, işlem o an <strong>login olduğunuz kullanıcı hesabı</strong> ile gerçekleştirilir. Bu hesabın ilgili kaynak üzerinde yetkisi yoksa işlem başarısız olur.</p><p>Daha da önemlisi, <strong>Zero Trust prensipleri gereği</strong>, günlük operasyonlarda yüksek yetkili kullanıcı hesapları kullanmaktan kaçınılmalıdır.</p><p>Bu noktada setAzureSPN scripti devreye girer:</p><ul><li>spn-automation adında bir <strong>Service Principal (SPN)</strong> oluşturur</li><li>Bu SPN’e ilgili subscription üzerinde <strong>Contributor</strong> rolü atar</li></ul><blockquote>(Bu örnekte Azure Storage üzerinden ilerliyoruz; ancak Key Vault, Redis ve benzeri servislerde gerekli rol atamaları farklılık gösterebilir.)</blockquote><p>💡 Not:<br>Firewall / networking ayarlarını yönetmek için <strong>Contributor rolü yeterlidir</strong>. Data-plane erişimi gerektiren senaryolar dışında ek roller (örneğin Storage Blob Data Contributor) gerekli değildir.</p><h3>⚠️ Önemli Not</h3><p>Bu script rol ataması yaptığı için, ilk çalıştırma sırasında <strong>Owner yetkisine sahip bir hesap</strong> gereklidir.</p><p>Ancak bu işlem:</p><ul><li>Sadece <strong>bir kez</strong> yapılır</li><li>Sonrasında tüm işlemler SPN üzerinden yürütülür</li><li>Ekstra yüksek yetkili kullanıcı kullanımına gerek kalmaz</li></ul><h3>🔐 Değişken Yönetimi</h3><p>Tenant ID, Client ID, Secret ve Subscription ID bilgileri local ortamda <strong>environment variable</strong> olarak saklanır.</p><p>Bu sayede:</p><ul><li>Bilgiler dış ortamlara gönderilmez</li><li>Ancak güvenli kullanım için doğru yapılandırma ve saklama yöntemleri uygulanmalıdır</li></ul><h3>⚙️ Nasıl Çalışır?</h3><p>Bu yaklaşım genellikle şu bileşenlerden oluşur:</p><ul><li>Basit bir <strong>Bash veya Python script</strong><br><em>(PowerShell ile de uygulanabilir; ancak Bash, özellikle Linux ortamlarında daha hafif ve esnek bir çalışma modeli sunar. Windows kullanıcıları için WSL güçlü bir alternatiftir.)</em></li><li><strong>Azure CLI (</strong><strong>az) entegrasyonu</strong></li><li><strong>Scheduler</strong><br><em>(cron, Azure Automation, GitHub Actions)</em></li></ul><p>Script düzenli aralıklarla çalışır ve erişim her zaman güncel IP üzerinden sağlanır.</p><h3>🔧 Kurulum ve Zamanlama</h3><p>Repository’yi klonladığınızı varsayıyorum:</p><pre>git clone https://github.com/gbutun/scripts.git</pre><h3>1️⃣ Script’e çalıştırma yetkisi verin</h3><pre>chmod +x /path/to/setAzureIP.sh</pre><h3>2️⃣ Crontab’ı açın</h3><pre>crontab -e</pre><h3>3️⃣ Aşağıdaki satırı ekleyin (her gün saat 01:00)</h3><pre>0 1 * * * /usr/bin/bash /path/to/setAzureIP.sh &gt;&gt; /var/log/setAzureIP.log 2&gt;&amp;1</pre><p>Bu şekilde script otomatik olarak çalışır, IP adresinizi günceller ve log’ları belirtilen dosyaya yazar.</p><h3>🎯 Kazanımlar</h3><p>Bu küçük otomasyonun etkisi düşündüğünüzden çok daha büyüktür:</p><p>✔ IP sprawl ortadan kalkar<br>✔ Saldırı yüzeyi kontrol altına alınır<br>✔ Manuel işlem ihtiyacı azalır<br>✔ Erişim modeli deterministik hale gelir<br>✔ Denetlenebilirlik (auditability) artar</p><h3>🧭 Stratejik Perspektif</h3><p>Bu yaklaşım sadece operasyonel bir kolaylık sağlamaz.</p><p>Aynı zamanda:</p><ul><li><strong>Zero Trust prensiplerine uyum sağlar</strong></li><li><strong>Least privilege erişim modelini destekler</strong></li><li><strong>Cloud governance olgunluğunu artırır</strong></li><li><strong>Operational excellence’a katkı sağlar</strong></li></ul><h3>🧩 Sonuç</h3><p>Bazen en kritik problemler, en basit görünen detayların içinde saklıdır.</p><p>Dynamic IP yönetimi de bunlardan biridir.</p><p>Doğru ele alınmadığında büyüyen bir risk,<br>doğru bir yaklaşımla kolayca kontrol altına alınabilir.</p><p><strong>Basit bir script, gerçekten büyük bir fark yaratabilir.</strong></p><p><strong>Sağlıcakla kalın ve bizi takip edin.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EvAaZrqsX43EkjBaKy7MtA.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dd5c57512999" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Azure VWAN route tables, route intent and policies, route maps.]]></title>
            <link>https://gbutun.medium.com/azure-vwan-route-tables-route-intent-and-policies-route-maps-d53eeaaef696?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/d53eeaaef696</guid>
            <category><![CDATA[vwan]]></category>
            <category><![CDATA[routing]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Thu, 27 Nov 2025 12:54:19 GMT</pubDate>
            <atom:updated>2025-11-27T12:54:19.000Z</atom:updated>
            <content:encoded><![CDATA[<h3>Azure VWAN route tables, route intent and policies, route maps. What are they, why is it important to know them. Part 3</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f6C7bGjs-jq7JDAPOsC3Ag.jpeg" /></figure><p>In this part we will be discussing route tables in VWAN. Like the name suggests, it is a table holding a route of a package from its coming source to its destination. In terms of VWAN, it is a logical container inside the vhub defining how traffic is forwarded between your connected Vnets, branches, and the Internet. It is essentially the policy engine that decides which next hop each packet should take.</p><p>As also mentioned in my first post, it has its kind of difficulties different than the traditional routing. To dive in we need to get familiar with some key concepts.</p><h3>Default vs Custom Route Tables</h3><ul><li>Default Route Table: Automatically created in every hub. All connections are associated with it unless you specify otherwise.</li><li>Custom Route Table: It is User-Defined. It used to enforce specific routing policies like forcing all internet traffic through and Azure Firewall which is quite a common scenario.</li></ul><h3>Association</h3><p>Now one of the concepts where things are getting interesting. Association with a connection like Vnet, Express Route or, VPN to a route table means that connection uses the routes in that table to forward the traffic. Let me clarify that, assume you have a VPN S2S connection to VWAN and if you associate that with the route table you basically say that the packages coming from that connection should use the route in the table to reach the destination.</p><p>Let us consume this with an example. Assume you have a VPN S2S connection into your VWAN. Associate this with a custom rote table having the default route 0.0.0.0/0 pointing it to the Azure Firewall. What happens is that all traffic arriving from that VPN site, regardless of its destination, is forwarded first to the Azure Firewall. The firewall becomes the mandatory inspection point, applying its configured network and application rules before allowing or denying the traffic. If permitted, the firewall then forwards the traffic to its intended destination, whether that is a spoke VNet, another on‑premises site via ExpressRoute, or the internet. This design enforces centralized security and ensures that every packet entering through the VPN connection is subject to consistent policy control and audit visibility.</p><h3>Propagation</h3><p>Another important piece of the topic. This is about a connection like VPN S2S, ExpressRoute, or User VPN (P2S) advertising its own routes into one or multiple route tables.</p><p>Let us check an example; Suppose you have a VPN S2S connection with on‑premises prefixes 192.168.10.0/24 and 192.168.1.0/24. If you configure propagation to the DefaultRouteTable, those prefixes appear in the table. Any VNet associated with the DefaultRouteTable can now send traffic to those on‑premises networks.</p><p>✅ Key Takeaway</p><p><strong>Propagation = advertise routes into a table.</strong></p><p><strong>Association = consume routes from a table.</strong></p><p>Together, they define how traffic flows in VWAN: who shares routes, and who can use them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oehooBK-yMPrxMSkKzsCTQ.png" /></figure><p><strong>How to Interpret Each Column</strong></p><p><strong>Prefix</strong></p><ul><li>The destination network (CIDR block) that this route applies to.</li><li>Example: 192.94.202.0/24 means traffic destined for that subnet.</li><li>0.0.0.0 is the default route, meaning “all destinations.”</li></ul><p><strong>Next Hop Type</strong></p><ul><li>The type of resource that traffic will be forwarded to.</li><li>Examples:</li><li><strong>- Azure Firewall </strong>→ traffic is inspected by the firewall.</li><li><strong>- Virtual Network Connection </strong>→ traffic goes to another VNet connected to the hub.</li><li><strong>- VPN_S2S_Gateway </strong>→ traffic is sent through the site-to-site VPN gateway.</li></ul><p><strong>Next Hop</strong></p><ul><li>The specific resource instance that traffic is forwarded to.</li><li>Example: n-cac-hub-fw-ip-cid is the firewall resource; n-cac-hub-vpngw-cid is the VPN gateway.</li></ul><p><strong>Origin</strong></p><ul><li>Where the route came from (which connection or resource propagated it).</li><li>Example: a VPN gateway connection propagates its on-premises prefixes into the table.</li></ul><p><strong>AS Path</strong></p><ul><li>Used in BGP routing. Shows the Autonomous System path the route traveled.</li><li>Empty here, meaning these are direct routes without AS path attributes.</li></ul><p><strong>🔎 Putting It Together</strong></p><ul><li>The <strong>default route (0.0.0.0)</strong> points to the <strong>Azure Firewall</strong>, so all traffic without a more specific match is inspected first.</li><li>The <strong>172.16.0.0/12</strong> and <strong>10.0.0.0/12</strong> prefixes are reachable via a <strong>Virtual Network Connection</strong>.</li><li>The <strong>192.94.202.0/24</strong>, <strong>207.142.98.0/22</strong>, and <strong>207.142.98.192/28</strong> prefixes are propagated from the <strong>VPN S2S Gateway</strong>, meaning they represent on‑premises networks advertised into the hub.</li></ul><p><strong>✅ Key Takeaway</strong></p><p>This table is essentially the <strong>routing map</strong> for your hub:</p><ul><li>It tells you which destinations are known.</li><li>It shows which connection advertised them (propagation).</li><li>It defines where traffic will be forwarded (association).</li></ul><h3>Labels</h3><p>Labels make it easier to organize and manage route tables. They are especially useful when you need to propagate routes from a connection to multiple route tables at once.</p><p>For example, every Default Route Table comes with a built-in label called “Default.” If you choose to propagate connection routes to the Default label, those routes will automatically be applied to all Default Route Tables across every hub in the virtual WAN.</p><p>If you don’t specify any label when setting up route propagation for a virtual network connection, the connection will automatically propagate to the Default label by default.</p><h3>Static routes in a virtual network connection</h3><p>Static routes let you control how traffic flows through the hub. By configuring a static route, you can direct traffic to a specific next hop IP address — for example, the IP of a Network Virtual Appliance (NVA) deployed in a spoke virtual network connected to the hub.</p><p>A static route consists of three key parts:</p><ul><li>Route name — the identifier for the route</li><li>Destination prefixes — the list of address ranges the route applies to</li><li>Next hop IP — the IP address where traffic should be forwarded</li></ul><h3>Considerations</h3><p>Key Points to Keep in Mind for Virtual WAN Routing</p><ul><li><strong>Branch connections and the Default route table<br> </strong>All branch connections (Point-to-site, Site-to-site, and ExpressRoute) must be linked to the <strong>Default route table</strong>. This ensures every branch learns the same prefixes.</li><li><strong>Consistent route propagation<br> </strong>Branch connections should propagate their routes to the same set of route tables. For example, if you choose the Default route table, apply this consistently across all branches. This way, all connections associated with the Default route table can reach each other.</li><li><strong>Azure Firewall across regions<br> </strong>If you deploy Azure Firewall in multiple regions, all spoke VNets must be tied to the same route table. Mixing VNets that go through the firewall with VNets that bypass it in the same hub isn’t supported.</li><li><strong>Multiple next hop IPs<br> </strong>A virtual network connection can have multiple next hop IPs. However, it cannot use multiple unique next hop IPs pointing to the same NVA in a spoke VNet if one of those routes is set to a public IP or 0.0.0.0/0 (internet).</li><li><strong>Internet routes are local only<br> </strong>Routes for 0.0.0.0/0 stay confined to the local hub’s route table. They do not propagate across hubs.</li><li><strong>Prefix rules for spoke routing<br> </strong>Virtual WAN can only program routes in a spoke if the prefix is broader (less specific) than the spoke’s VNet prefix. For example, if a spoke VNet has 10.1.0.0/16, Virtual WAN cannot inject routes for 10.1.0.0/16 or its subnets like 10.1.0.0/24.</li><li><strong>Dynamic vs. static routing between hubs<br> </strong>Hubs in the same Virtual WAN share routes dynamically (if propagation is enabled to the same labels). Static routes, however, do not get shared between hubs.</li><li><strong>Avoid hub router IPs as next hop<br> </strong>When defining static routes, do not use hub router IPs as the next hop.</li><li><strong>BGP community removal caution<br> </strong>If you strip Azure BGP communities from VNet or UDR routes, do not advertise those routes back into Virtual WAN. Doing so can cause routing problems and outages.</li></ul><p>In these parts we have seen the basics of routing. When we sum together VWAN as a concept seems to be with good intentions but has some flaws like complex management or hidden un supported scenarios.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d53eeaaef696" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Azure VWAN route tables, route intent and policies, route maps.]]></title>
            <link>https://gbutun.medium.com/azure-vwan-route-tables-route-intent-and-policies-route-maps-c4a84ea9ba3d?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/c4a84ea9ba3d</guid>
            <category><![CDATA[azure-vwan]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Fri, 31 Oct 2025 12:20:36 GMT</pubDate>
            <atom:updated>2025-10-31T12:20:36.074Z</atom:updated>
            <content:encoded><![CDATA[<h3>Azure VWAN route tables, route intent and policies, route maps. What are they, why is it important to know them. Part 2</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J8OdPaUqW3n8XYS8qzpgDA.jpeg" /></figure><p>In this second part we will check on routing Intent and policies.</p><p>What is it? If you are dealing with Azure VWAN, this is one of the most confusing topics. Not that is difficult, but more the question is would be why and when we have custom routes.</p><p>Basically It allows you to configure Virtual WAN hub to send Internet-bound and Private (Point-to-site, Site-to-site, ExpressRoute, Network Virtual Appliances inside the Virtual WAN Hub and virtual network) Traffic via an Azure Firewall, Next-Generation Firewall NVA or software-as-a-service solution deployed in the Virtual WAN hub.</p><h4>A question to ask why use Hub routing intent and policies over custom route tables?</h4><p>It is easy, we have just two options here in a declarative way :</p><ul><li><strong>Internet Traffic Policy →</strong> sends all Internet‑bound traffic from VNets, branches, and P2S/ER/S2S connections to a chosen next hop (Azure Firewall, NVA, SaaS security provider).</li><li><strong>Private Traffic Policy →</strong> forces all private traffic (VNet ↔ branch, branch ↔ branch, inter‑hub) through the chosen next hop.</li><li><strong>Automatic propagation</strong></li><li><strong>Consistent</strong></li><li><strong>Integration with the Secure Virtual Hub</strong></li></ul><p>Assume you do not need too much, should be quick and easy.</p><h3>Prerequisites</h3><p><strong>✅ Virtual Hub Requirements</strong></p><h4>Your Virtual Hub must meet <strong>all</strong> of the following conditions before you can enable routing intent:</h4><p><strong>No custom route tables</strong></p><ul><li>Only the following route tables may exist:</li><li>— noneRouteTable</li><li>— defaultRouteTable</li></ul><p><strong>No static routes with next hop Virtual Network Connection</strong></p><ul><li>— Static routes <strong>are allowed</strong> in the defaultRouteTable if the next hop is <strong>Azure Firewall</strong>.</li></ul><p><strong>Configuration option availability</strong></p><ul><li>— If the above requirements are not met, the <strong>Routing Intent option will be greyed out</strong> in the portal.</li></ul><p><strong>🔄 Additional Requirement (Azure Firewall Manager — Inter-Hub Option)</strong></p><p>When using <strong>Routing Intent with the inter-hub option</strong> in <strong>Azure Firewall Manager</strong>, there is an extra constraint:</p><ul><li>Routes created by Azure Firewall Manager must follow the naming convention:<br> — private_traffic<br> — internet_traffic<br> — all_traffic</li><li>Therefore, <strong>all routes in the defaultRouteTable must also follow this convention</strong>.</li></ul><h3>When we look to custom route tables we have advantages like:</h3><ul><li>Granular control: You can create multiple route tables per hub and associate/propagate connections selectively.</li><li>Flexibility: Useful for advanced scenarios like multi‑tenant isolation, overlapping address spaces, or custom east‑west segmentation.</li><li>Manual effort: You must explicitly configure associations and propagations, and maintain them as the environment grows.</li><li>Risk of drift: Easy to misconfigure, leading to asymmetric routing or bypassing security appliances.</li></ul><p>Especially good for advanced scenarios or multi-tenant hosting scenarios, route Intent will fail here. Also you cannot use both at the same time.</p><h3>Below a matrix which I compiled with help of copilot but true :)</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k4CL84VfKkueRhpF8NIe3A.png" /></figure><p>Now the most pressing question would be to use it where, what are the use cases? There are a few you check under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies#use-cases">use cases</a> . Since repeating them here doesn’t make sense I just put the link.</p><p>There are some points you need to keep in mind. These are:</p><ul><li>Routing intent can only be configured on hubs that do not contain custom route tables and where the default route table has no static routes with a next hop set to a Virtual Network Connection. This can be verified easily, setup route intent and try to create a custom route table, you will see you cannot. More information is under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies#prereq">prerequisites</a>.</li><li>A copy of gateways, connections, and route tables should be saved before enabling routing intent, as the system does not automatically preserve or reapply previous configurations. More information is under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies#rollback">rollback strategy</a>.</li><li>When routing intent is enabled, static routes in the defaultRouteTable are modified. Because of Azure portal optimizations, the resulting state of the defaultRouteTable may differ from configurations applied through REST, CLI, or PowerShell. More information is under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies#staticroute">static routes</a>.</li><li>When routing intent is enabled, the set of prefixes advertised to on‑premises networks is modified. More information is under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies#prefixadvertisments">prefix advertisements</a></li><li>A support case can be opened to enable connectivity across ExpressRoute circuits through a Firewall appliance in the hub. Enabling this connectivity pattern changes the prefixes that are advertised to ExpressRoute circuits. More information is under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies#expressroute">About ExpressRoute</a></li><li>In Virtual WAN, routing intent is the sole mechanism that enables inter‑hub traffic inspection through security appliances deployed in the hubs. For consistent and symmetrical routing of inter‑hub traffic between these appliances, routing intent must be enabled on all hubs.</li><li>Routing intent directs Virtual Network and on‑premises traffic to the next hop resource defined in the routing policy. In this model, the Azure platform handles the routing according to the configured policy, bypassing the Virtual Hub router. Since packets are not processed by the hub router, additional routing infrastructure units are not required for data‑plane packet forwarding on hubs with routing intent enabled. However, extra routing infrastructure units may still be needed depending on the number of Virtual Machines connected to Virtual Networks within the Virtual WAN Hub.</li><li>Routing intent supports configuring different next‑hop resources for private and internet routing policies. For instance, the next hop for private routing can be set to Azure Firewall in the hub, while the next hop for internet routing can be directed to an NVA or a SaaS solution in the hub. Since SaaS solutions and Firewall NVAs share the same subnet within a Virtual WAN hub, deploying both in the same hub can limit the horizontal scalability of the SaaS solution due to reduced IP address availability. In addition, only one SaaS solution can be deployed per Virtual WAN hub.</li></ul><h4>Limitations</h4><p><strong>🌍 Availability by Cloud Environment</strong></p><ul><li><strong>Routing intent is not available</strong> in Microsoft Azure operated by 21Vianet.</li><li><strong>Palo Alto Cloud NGFW</strong> is only available in <strong>Azure Public</strong>.</li><li>For Azure Government or 21Vianet, contact Palo Alto Networks directly.</li><li><strong>Network Virtual Appliances (NVAs)</strong> are not available in all Azure Government regions.</li><li>Check with your NVA partner for region-specific availability.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/977/1*0zE3HUSRoV2YwYlhBZl57A.png" /></figure><p><strong>🔧 What Routing Intent Does</strong></p><p>Simplifies routing by <strong>managing route table associations and propagations</strong> for all connections:</p><ul><li>Virtual Network</li><li>Site-to-Site VPN</li><li>Point-to-Site VPN</li><li>ExpressRoute</li></ul><p><strong>Not compatible</strong> with Virtual WANs that use <strong>custom route tables or customized policies</strong>.</p><p><strong>🔐 Encrypted ExpressRoute Support</strong></p><ul><li>Supported in hubs where <strong>routing intent is configured</strong>.</li><li>Requires <strong>Azure Firewall</strong> to allow traffic between:<br> — Site-to-Site VPN Gateway private IP<br> — On-premises VPN device private IP</li></ul><blockquote>See: <em>Encrypted ExpressRoute with Routing Intent</em> for configuration details in the official documentation.</blockquote><p><strong>🚫 Unsupported Connectivity Scenarios</strong></p><p><strong>Static routes</strong> in defaultRouteTable pointing to a Virtual Network connection.</p><ul><li>Use <strong>BGP peering</strong> instead.</li></ul><p><strong>Static route propagation</strong> on Virtual Network connections.</p><ul><li>Not applied to private routing policy next-hop (on roadmap).</li></ul><p><strong>SD-WAN NVA + Firewall NVA/SaaS in the same hub</strong>.</p><ul><li>Causes connectivity issues.</li><li>Workarounds:<br> — Deploy in <strong>different Virtual Hubs</strong>, or<br> — Place SD-WAN NVA in a <strong>spoke VNet</strong> and use <strong>hub BGP peering</strong>.</li></ul><p><strong>🔥 Network Virtual Appliance (NVA) Restrictions</strong></p><ul><li>Only <strong>Next-Generation Firewalls (NGFWs)</strong> or <strong>dual-role NGFW + SD-WAN NVAs</strong> are supported.</li><li>Eligible NVAs:<br>- checkpoint<br>-fortinet-ngfw<br>-fortinet-ngfw-and-sdwan<br>-cisco-tdv-vwan-nva</li><li><strong>Palo Alto Cloud NGFW</strong> is supported as a <strong>SaaS next hop</strong>, not as an NVA.<br>-If you specify an unsupported NVA, <strong>Routing Intent creation fails</strong>.<br>-To verify: <br> -Navigate to <strong>Virtual Hub → Network Virtual Appliances → Vendor field</strong>.</li></ul><p><strong>🔄 ExpressRoute Circuit Connectivity</strong></p><ul><li>To connect <strong>multiple ExpressRoute circuits</strong> through a security solution in the hub:</li><li>Open a <strong>support case</strong> to enable this scenario.</li></ul><blockquote>See: <em>Enabling connectivity across ExpressRoute circuits in thg official document.</em></blockquote><p>✨ This structure makes it <strong>role-aware and scannable</strong>:</p><ul><li><strong>Execs/Managers</strong> can stop at the availability table.</li><li><strong>Engineers</strong> can dive into unsupported scenarios and NVA restrictions.</li><li><strong>Auditors</strong> can trace limitations and roadmap items clearly.</li></ul><p>Routing Intent and policies is a quite vast topic with a lot of things like Transit connectivity between ExpressRoute circuits or how it is encrypted. Since our aim is to give a taste of it you can follow it under <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/how-to-routing-policies">How to configure Virtual WAN Hub routing intent and routing policies</a></p><p>In the upcoming part we will discuss more features of VWAN demystifying them, stay safe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4a84ea9ba3d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Azure VWAN route tables, route intent and policies, route maps.]]></title>
            <link>https://gbutun.medium.com/azure-vwan-route-tables-route-intent-and-policies-route-maps-736d4503a760?source=rss-08e3ce050892------2</link>
            <guid isPermaLink="false">https://medium.com/p/736d4503a760</guid>
            <category><![CDATA[azure-networking]]></category>
            <category><![CDATA[azure-vwan]]></category>
            <dc:creator><![CDATA[Gultekin Butun]]></dc:creator>
            <pubDate>Fri, 17 Oct 2025 11:07:12 GMT</pubDate>
            <atom:updated>2025-10-17T11:07:12.512Z</atom:updated>
            <content:encoded><![CDATA[<h3>Azure VWAN route tables, route intent and policies, route maps. What are they, why is it important to know them. Part 1</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f6C7bGjs-jq7JDAPOsC3Ag.jpeg" /></figure><p>According to Microsoft documentation VWAN or Virtual WAN is “a networking service that brings many networking, security, and routing functionalities together to provide a single operational interface. “</p><p><strong>Some of the main features include:</strong></p><ul><li>Branch connectivity (via connectivity automation from Virtual WAN Partner devices such as SD-WAN or VPN CPE).</li><li>Site-to-site VPN connectivity.</li><li>Remote user VPN connectivity (point-to-site).</li><li>Private connectivity (ExpressRoute).</li><li>Intra-cloud connectivity (transitive connectivity for virtual networks).</li><li>VPN ExpressRoute inter-connectivity.</li><li>Routing, Azure Firewall, and encryption for private connectivity.</li></ul><p>Without AI/Copilot this means as follows; VWAN is a Microsoft managed networking structure perfect for companies which has a small team or doesn’t want to deal with it.</p><p>Since it is managed by Microsoft there are a lot of advantages like routing, transit connectivity, easier firewall or NVA integration (natively integration with less hassle), automatically meshing, centralized management of route tables, easier to audit, scaled very well and has better limits, native HA capabilities.</p><p>There are a lot of advantages but now the reality; even though it has many features, the configuration at the beginning is not as easy as advertised. I believe this is the main reason why VWAN is not that popular and let you rethink what you know. For this reason I believe that this post would be useful.</p><p>As usual in every post let us start with some terms and concepts; what they mean and under which condition to use.</p><h3>Route Maps</h3><p>Per official documentation the definition is as, “a feature that gives you the ability to control route advertisements and routing for Virtual WAN virtual hubs.”</p><p>This means when there are routes that are advertised (send to Azure), we have the option to control it. We can accept or denied it and can do more of it.</p><p><strong>Some benefits which are in the official document but better explained if you will:</strong></p><ul><li>In connecting to remote networks from VWAN there can be a great number of exchanged routes which can hit limits or can be overwhelming. This is especially the case with BGP, even a great protocol but comes with a price.</li></ul><p>An example would be; let say we have the following advertised,</p><p><em>10.122.1.0/24, 10.122.2.0/24, 10.122.3.0/24</em>. This can be aggregated to <em>10.122.0.0/16</em>. Surely this is a very simple example, especially advertisements from a big environment result in pretty big routes but route-map has it under control.</p><ul><li>Control is everything. How do we control incoming and outgoing to our VWAN? A very important feature which is very much related to security. Imagine you have a multitenant application hosted on Azure, you need to isolate every bit of information so they are not reached by another customer you have or vice versa, you do not want Azure to learn about some resources in another subnet. Let say the remote environment or on-premises advertises 10.0.0.0/8 but this is too much and you just want 10.0.0.0/16.</li></ul><p>We have discussed more incoming but how about outgoing? The VHUB also advertises which means the remote tenant or on-premises is also being advertised. We can filter them as well and we should especially in multitenant environments. Assume you have 100 VNET prefixes but you wish advertise a summarized 10.122.0.0/16 or want prevent certain VNETs being reachable from remote or on premises.</p><p>Controlling routing decisions by changing the BGP attribute as AS-PATH to make a route more, or less preferable. So what does this mean; this requires some background in BGP. When we have two or more routes having the same prefix, the router by de design must pick up the best path. So what is the best path?</p><ul><li>In Azure VWAN (mostly anywhere I know) the default behavior is like, longest prefix match → static routes → routing preference → BGP tie‑breakers. So the shorter AS-PATH is preferred over the longer. Let us explain this with a good example. Assume we have a prefix of 10.10.0.0/16 and can be reached via ER and VPN. Depending on the routing preference (the routing preference can be set in portal or with other methods), VWAN can pick up both. With the AS-PATH prepend in Route-Map on the VPN (65001) the hub sees the path longer and prefers the ER. If ER fails VPN is picked up.</li></ul><p>This helpful in the decision of primary and secondary connections.</p><ul><li>Tagging routes using the BGP community feature to manage routes. To get deep we need explain what the community tag is. It is a metadata attached to a BGP route or prefix if you will. It is about adding information so in VWAN allowing enterprises to classify and control routes.</li></ul><p>When we wrapped it up, route-maps do:</p><ul><li><strong>Summarize</strong> → reduce many prefixes into one.</li><li><strong>Filter </strong>→ block specific prefixes from being advertised.</li><li><strong>Modify BGP attributes </strong>→ e.g., prepend AS‑PATH to make one path less preferred, or tag with BGP communities.</li><li><strong>Selective reachability</strong> → decide which VNets are visible to which branches/ER circuits.</li></ul><h4>Limitations</h4><p>Although it has a lot of cool features, it has also some limitations. Below the most important ones, since it does not make sense to repeat them you can find them under the link <a href="https://learn.microsoft.com/en-us/azure/virtual-wan/route-maps-about#considerations-and-limitations">https://learn.microsoft.com/en-us/azure/virtual-wan/route-maps-about#considerations-and-limitations</a></p><ul><li>When summarizing the routes with route-maps the hub router strips the BGP Community and AS-PATH attributes from those routes. It applies to both in and outbound routes. Now this seems to be contradicting with the tagging feature of route-maps. In fact what happens is route-map generate a new aggregated route so it does not make sense to keep the previous ones since they do not exists any more. From AS-PATH point of view, we aggregate multiple routes with different AS-PATHs since there is no single AS-PATH to attach, to avoid any ambiguity it is striped. We can say the same for the community field, assume you have multiple community fields before so which one to inherit?</li><li>No usage of private ASN numbers, there is a great article on this. <a href="https://learn.microsoft.com/en-us/azure/expressroute/expressroute-optimize-routing">do not use private ASNs for AS prepending</a></li><li>No usage of ASN’s reserved by Azure for AS.</li><li>Public ASNs: 8074, 8075, 12076</li><li>Private ASNs: 65515, 65517, 65518, 65519, 65520</li><li>You can’t apply Route-maps to connections <strong>between on-premises and SD-WAN/Firewall NVAs</strong> in the virtual hub. You can still apply route-maps to other supported connections when an NVA in the virtual hub is deployed. <strong>This doesn’t apply to the Azure Firewall</strong>, as the routing for Azure Firewall is provided through Virtual WAN routing intent features.</li></ul><blockquote>Note: Since I did not give the whole list, it makes sense to check them in alignment with your architecture.</blockquote><h3><a href="https://learn.microsoft.com/en-us/azure/virtual-wan/route-maps-how-to">How to configure Route-maps</a></h3><p>The link speaks for itself, for clarity I will break it down a bit.</p><p><strong>Some important points if you want to use route-maps:</strong></p><ul><li>It is not free; see the <a href="https://azure.microsoft.com/pricing/details/virtual-wan/">Pricing</a> page.</li><li>Can be applied to S2S, ER or P2S.</li><li>Check your design, do you really need it? It has a price and can be complicated with its rules.</li></ul><p>In the upcoming part we will discuss more features of VWAN demystifying them, stay safe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=736d4503a760" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>