Quantcast
Channel: TechNet Blogs
Viewing all 36188 articles
Browse latest View live

Windows Defender ATP の機械学習と Antimalware Scan Interface: スクリプトを悪用した「環境寄生型」攻撃の検出

$
0
0

本記事は、Microsoft Secure のブログ “Windows Defender ATP machine learning and AMSI: Unearthing script-based attacks that ‘live off the land’” (2017 12 4 日 米国時間公開) を翻訳したものです。


高度な標的型攻撃を実行するアクティビティ グループや、無差別型の脅威を展開するマルウェア作成者によるスクリプトの悪用が進んでいます。

JavaScript、VBScript、PowerShell などのスクリプト エンジンは、攻撃者にとって大きなメリットがあります。これらは正規のプロセスを通じて実行されるため、「環境寄生型」攻撃 (ディスクを操作するのではなく、一般的なツールを使用してインメモリでコードを直接実行する攻撃 (英語)) のためのツールとして最適です。スクリプト エンジンの多くには、オペレーティング システムの一部として、インターネット上のコンテンツをオンザフライで評価して実行する機能が備わっています。さらに、広く使用されているアプリと統合すれば、ソーシャル エンジニアリングを通じて悪意のあるコンテンツを配信するうえで効果的な手段にもなります。この点は、スパム攻撃におけるスクリプトの使用 (英語) が増加していることからも明らかです。

悪意のあるスクリプトは、配信メカニズムとしてだけでなく、キル チェーンのさまざまな段階で使用されます。たとえば、感染後の被害拡大や持続性の確保などです。これらの段階では、使用するスクリプト エンジンとして PowerShell が選ばれることが圧倒的に多くなります。PowerShell は Windows で管理タスクを実行するためのスクリプト エンジンのデファクト スタンダードであり、システム API を呼び出すことができるほか、幅広いシステム クラスやオブジェクトにアクセスできるためです。

スクリプトは、強力なスクリプト エンジンを利用することで便利なツールにもなりますが、その動的な性質から、マルウェア対策製品や同様のエンドポイント保護製品による分析や検出を回避しやすくするための手段として、攻撃者に悪用される場合もあります。難読化が容易なうえに、リモート サイトやレジストリ キーからオンデマンドで読み込むことが可能なスクリプトは、検出がきわめて困難です。

Windows 10 では、Antimalware Scan Interface (AMSI、英語) を通じてスクリプトの動作に関するインサイトが提供されます。AMSI は、Windows Defender ウイルス対策がスクリプト インタープリターと同様に、暗号化と難読化の両方を解除した状態でスクリプトの内容を検査できるようにする汎用オープン インターフェイスです。Windows 10 Fall Creators Update (英語) では、長年にわたるスクリプト ベースのマルウェアの分析から得られた知識に基づいて、Windows のスクリプト インタープリター自体に高度な動作インストルメンテーションを追加したことで、スクリプトによるシステム操作を取得できるようになりました。こうした操作に関する詳細情報を、Windows Defender ウイルス対策などの登録済みの AMSI プロバイダーで利用し、ランタイム スクリプト実行コンテンツのさらなる検査と検証を実行できます。

スクリプトの動作の可視化という他に類を見ない機能は、Windows 10 Fall Creators Update における他の機能強化 (Windows Defender ウイルス対策および Windows Defender Advanced Threat Protection (Windows Defender ATP)) でも活用されています。いずれのソリューションも、強化されたインサイトを処理する強力な機械学習アルゴリズムを活用しており、Windows Defender ウイルス対策では悪意のあるスクリプトを侵入前にブロックする機能が強化され、Windows Defender ATP では悪意のあるスクリプトの侵入後の行動に基づいて効果的にアラートを生成する機能が追加されました。

今回のブログでは、特に Windows Defender ATP において、AMSI の検査データを利用して、回避行動を取る複雑なスクリプト ベースの攻撃を検出するしくみについて説明します。高度なスキルを持つアクティビティ グループの KRYPTON が実行した高度な攻撃を取り上げるほか、Kovter などの無差別型のマルウェアが PowerShell を悪用し、ディスク上に悪意のあるアクティビティの痕跡をほとんど残さずに活動する方法について説明します。さらに、Windows Defender ATP の機械学習システムでスクリプトの特性や動作に関する強化されたインサイトを利用することで、検出機能を大幅に強化する方法についても説明します。

KRYPTON: スクリプト ベースの攻撃の耐久性

従来、潜在的なセキュリティ侵害を検出する方法の大半はファイルを調査するものでした。インシデント対応担当者は多くの場合、自動開始エントリをトリアージし、普及率やファイル名とフォルダーの不自然な組み合わせなどから不審なファイルを特定します。しかし、最新の攻撃は完全なファイルレスの手法へと進化しつつあり、適切なチョーク ポイントにセンサーを追加する必要に迫られています。

最新のスクリプト ベースの攻撃では、ディスク上のファイルを利用しないだけでなく、多くの場合、暗号化された悪意のあるコンテンツと復号化キーが別々に格納されます。さらに、最終的なキーを使用して実際のペイロードの暗号化を解除するまでには複数のプロセスが実行される場合が多く、実際にスクリプトの呼び出しを追跡せずに、1 つのファイルだけで判断することはできません。このタスクは、完全なスクリプト エミュレーターでも対応不可能です。

たとえば、アクティビティ グループの KRYPTON は、スケジュールされたタスクをハイジャックしたり作成したりしていることが確認されており、Autoruns for Windows (英語) などの一般的なフォレンジック ツールの実行リストに含まれるシステム タスクを標的とすることがよくあります。KRYPTON は、スケジュールされたタスクのパラメーター内に一意の復号化キーを格納し、実際のペイロードのコンテンツは暗号化された状態で保持されます。

KRYPTON の攻撃を説明するため、John LambertOffice 365 Advanced Threat Protection チームが特定した感染済みの Microsoft Word ドキュメントを紹介します。

図 1. KRYPTON の誘導ドキュメント

KRYPTON は環境寄生型の攻撃を行うので、マルウェア対策のアラートをトリガーするような従来型の悪意のあるバイナリを投下したり送り込んだりすることはありません。その代わりに、誘導ドキュメントにマクロが含まれており、Windows Scripting Host (wscript.exe) を使用して JavaScript ペイロードを実行します。このスクリプト ペイロードを実行するには適切な RC4 復号化キーが必要で、このキーは案の定、スケジュールされたタスクの引数として格納されています。このスクリプト ペイロードは、適切なキーを適切な順序で生成しなければトリガーされないため、サンドボックス環境での自動デトネーションや手動での検査に対する耐久性があります。

図 2. wscript.exe を使用した KRYPTON のスクリプト実行チェーン

AMSI による実際のスクリプトの動作の解明

AMSI は、KRYPTON の回避メカニズムを打破するために、復号化されてスクリプト インタープリターで実行する準備が整った後に JavaScript API の呼び出しを取得します。次のスクリーンショットは、AMSI によって取得された KRYPTON の攻撃から解明されたコンテンツの一部です。

図 3. AMSI によって取得され、分析用にクラウドに送信された KRYPTON のスクリプト ペイロードの一部

Windows Defender ATP は、取得したスクリプトの動作をセキュリティ専門家や機械学習アルゴリズムによって蓄積された攻撃の痕跡 (IoA: Indicator of Attack) に照らし合わせて確認することで、KRYPTON のスクリプトに悪意があることを難なく判定します。同時に、悪意のある Word ドキュメントからスクリプトがトリガーされたしくみなど、有意義なコンテキスト情報も提供します。

図 4. AMSI によって取得された KRYPTON のスクリプトを Windows Defender ATP の機械学習で検出

Kovter などの無差別型マルウェアによる PowerShell の使用

KRYPTON などの高度なアクティビティ グループ以外でも、バイナリ実行可能ファイルから、回避行動を取るスクリプトへの移行が進んでいます。無差別型マルウェアの分野では、Kovter (英語) が複数のプロセスを使用して、最終的に悪意のあるペイロードを実行しています。このペイロードは、(wscript.exe によって実行される) JavaScript で暗号化が解除される PowerShell スクリプト内に寄生し、環境変数として powershell.exe に渡されます。

図 5. Kovter のスクリプト ベースのペイロードの実行に対して Windows Defender ATP の機械学習でアラートを生成

AMSI によって取得された PowerShell ペイロードのコンテンツ (英語) を熟練のアナリストが確認すれば、一般公開されている侵入テスト用モジュール PowerSploit (英語) との類似性に容易に気付きます。このような攻撃手法にはファイル ベースのコンポーネントが使用されますが、悪意のあるアクティビティはインメモリのみで実行されるため、従来の手法で検出することは著しく困難です。しかし、Windows Defender ATP では機械学習を使用して、詳細な AMSI シグナルと一般的な PowerShell アクティビティによって生成されるシグナルを組み合わせることで、このような行動を容易に検出できます。

図 6. AMSI によって取得され、分析用にクラウドに送信された Kovter のスクリプト ペイロードの一部

AMSI で提供される機械学習を活用した新しいインサイト

AMSI は、取得したスクリプトのコンテンツから豊富な情報を提供しますが、悪意のあるスクリプトには亜種が非常に多く、検出は簡単ではありません。悪意のあるスクリプトと無害なスクリプトを区別する新たな特徴を効率的に抽出して特定するために、Windows Defender ATP では高度な機械学習の手法を採用しています。

こちらのブログ記事で説明したように、侵入アクティビティの特定には教師あり機械学習の分類子を使用します。マイクロソフトは、実際に確認された悪意のある行動と一般的なマシンの正常なアクティビティに基づいてトレーニング セットを構築し、制御下における悪意のある成果物のデトネーションから取得したデータを補完しています。次の図は、悪意のある行動を取得する方法をプロセス ツリーの形で概念的に示したものです。

図 7. AMSI データのインストルメンテーションによって強化されたプロセス ツリー

このプロセス ツリーにあるとおり、キル チェーンは悪意のあるドキュメントから始まり、Microsoft Word (winword.exe) によって PowerShell (powershell.exe) が起動されます。その後、高度に難読化されたスクリプトが PowerShell によって実行されます。このスクリプトは、fhjUQ72.tmp というマルウェアを投下および実行し、レジストリに Run キーを追加して持続性を確保します。マイクロソフトの機械学習システムは、このプロセス ツリーからさまざまな特徴を抽出し、レジストリ改変やファイル作成といった分野の「専用の分類子」を作成することができます。この分類子は数値のスコアに変換され、アラートを生成するかどうかの判断に使用されます。

Windows 10 Fall Creators Update (version 1709、英語) では AMSI シグナルのインストルメンテーションが追加されました。これにより、Windows Defender ATP の機械学習アルゴリズムは、難読化が解除されたスクリプトのコンテンツの分析に関するインサイトを利用すると共に、プロセスのアクティビティに関連したマシンの状態変化を継続的に参照できるようになりました。また、難読化層の数、エントロピ、難読化機能、n-grams、特定の API 呼び出しなど、実行されたスクリプトの性質を検査するさまざまなスクリプト ベースのモデルも構築されました。

AMSI で難読化層を順次解除することで、Windows Defender ATP には悪意のあるスクリプトの API 呼び出し、変数名、一般的な構造パターンに関するインサイトが蓄積され、可視性が高まります。さらに、AMSI のデータによってセキュリティ専門家の知識や機械学習システムのトレーニング能力が向上すると同時に、マイクロソフトのディープ ニューラル ネットワークでは人間のアナリストが発見できないような特徴を自動的に学習します。

図 8. JavaScript および PowerShell スクリプトを機械学習で検出

こうした新しいスクリプト ベースの機械学習モデルによって専門の分類子が強化される一方で、新たな結果と他の動作に関する情報の関連付けも行っています。たとえば、Windows Defender ATP では、AMSI で検出された悪意のあるスクリプトのコンテンツと、ネットワーク接続などの他の直近の動作が関連付けられます。このコンテキスト情報はセキュリティ運用担当者に提供され、効果的なインシデント対応に活用できます。

図 9. AMSI で検出された VBScript のコンテンツと追跡されたネットワーク アクティビティを機械学習で関連付け

AMSI をバイパスしようとする行動の検出

AMSI は悪意のあるスクリプトのアクティビティに関する強力なインサイトを提供するため、今後の攻撃には AMSI をバイパスするメカニズムが実装される可能性が高くなります。マイクロソフトは、これらのメカニズムを以下の 3 種類に分類しています。

  • スクリプトのコンテンツに組み込まれたバイパス (検査やアラートによって対応可能)
  • AMSI センサー インフラストラクチャの改ざん (システム ファイルの置換や関連 DLL ファイルの読み込み順序の操作など)
  • インメモリの AMSI インストルメンテーションへのパッチ適用

Windows Defender ATP 研究チームは、すべてのセンサーの改ざん対策メカニズムをプロアクティブに開発しています。潜在的なインサイトの操作に対応するヒューリスティック アラートを考案し、バイパスされる前にクラウドでアラートをトリガーするように設計しました。

実際に CVE-2017-8759 の攻撃が発生した際、Windows Defender ATP はエクスプロイト後の悪意のあるスクリプトのアクティビティを検出しただけでなく、Matt Graeber が特定したものと類似のコードを使用して AMSI をバイパスしようとしていることも検出しました。

図 10. AMSI をバイパスするパターンに基づく Windows Defender ATP のアラート

以下のバイパス コードは AMSI 自体によって取得され、Windows Defender ATP クラウドでの分析に使用されました。

図 11. 分析用にクラウドに送信された AMSI をバイパスしようとするコード

まとめ: Windows Defender ATP の機械学習と AMSI により、高度な回避行動を取るスクリプト ベースの攻撃に対する画期的な防御を実現

Windows 10 のオープン インターフェイスとして提供される Antimalware Scan Interface は、暗号化および難読化されたスクリプトに潜み、ディスクに書き込まれることのない悪意のあるアクティビティに関する強力なインサイトを提供します。このような回避行動を取るスクリプトは一般化しつつあり、高度なスキルを持つアクティビティ グループと無差別型マルウェアの作成者の両方が採用しています。

AMSI では、悪意のあるスクリプトの行動を取得するために、スクリプトのコンテンツの解釈と並行して検査を実行します。そのため、物理ファイルを確認する必要はなく、難読化、暗号化、ポリモーフィズムによって妨害されることもありません。エンドポイントでは、必要なインサイトを AMSI からローカル スキャナーに提供することで、難読化や暗号化が適用されているスクリプトでも悪意のあるコンテンツを検査することができます。特に Windows Defender ウイルス対策は、AMSI を利用して、ランサムウェアやバンキングを標的としたトロイの木馬など、あらゆる種類の悪意のあるペイロードを投下するスクリプトを動的に検査、ブロックします。

Windows 10 Fall Creators Update (1709、英語) では、新たに追加されたスクリプト ランタイム インストルメンテーションにより、難読化されているスクリプトの動作を可視化する他に類を見ない機能が提供されます。Windows Defender ウイルス対策では、この悪意のあるスクリプトの動作に関する豊富な情報を利用して、侵入前の保護をランタイムに提供します。さらに、侵入後の保護を提供するため、高度な機械学習システムを使用して、このデータから詳細なインサイトを引き出します。

Windows Defender ATP の新しい機械学習アルゴリズムでは、特定のアクティビティやアクティビティのパターンを確認するだけでなく、スクリプトの難読化層や API 呼び出しのパターンなど、さまざまな特徴を調査し、悪意のあるスクリプトをヒューリスティックな手法で効率的に特定するために使用しています。また、スクリプト ベースのインジケーターと他の直近のアクティビティを関連付けることで、侵入の疑いに関する豊富なコンテキスト情報を提供します。

新たに追加されたスクリプト ランタイム インストルメンテーションや Windows Defender Exploit Guard などの強力なセキュリティ機能をご利用になるために、Windows 10 Fall Creators Update (英語) をインストールすることをお勧めします。

Windows Defender ATP によって実現される大幅なコスト削減やビジネス上のメリットの詳細については、Forrester による Microsoft Windows Defender Advanced Threat Protection の Total Economic Impact レポート (英語) をご覧ください。Windows Defender ATP を使用して、お客様の企業で高度な攻撃の検出、調査、対処を実際にお試しになるには、無料試用版へのサインアップをお願いいたします。

 

Stefan Sellmer (Windows Defender ATP 研究チーム)

共同執筆:

Shay Kels (Windows Defender ATP 研究チーム)

Karthik Selvaraj (Windows Defender 研究チーム)

 

関連情報

 


Using Hashtables in your SCOM scripts

$
0
0

 

image

 

When we work with getting data out of SCOM via PowerShell, or building automations, something that can really help speed things up are Hashtables.

 

I was recently working with a customer to build a discovery for a new class, and get properties from other existing classes.  I had to build an array of these class objects (Windows Computers) and loop through them each time I wanted to use one.  As I found out – this became VERY time consuming.

In this case, I was doing something basic like this:

$WCarr = Get-SCOMClass -Name "Microsoft.Windows.Computer" | Get-SCOMClassInstance

That creates an array ($WCarr) and fills it with all the Windows Computer objects in the management group. 

The customer has about 10,000 Windows Computers in this SCOM Management group, so this operation takes about 85 seconds to complete.  Not a big deal for my script.

 

However, in a subsequent part of the script, I needed to loop through about 600 objects, and on each one I needed to get properties from the Windows Computer object.  Here is an example:

FOREACH ($WC in $WCarr)
{
     IF ($NameImLookingFor -eq $WC.DisplayName)
     {
          $MatchingComputerObject = $WC
     }
}

A cleaner way to write this same thing would be:

$MatchingComputerObject = $WCarr | where {$_.DisplayName -eq $NameImLookingFor}

 

I am basically just going through all 10,000 records in memory, until I find the one I am looking for.  At the customer, this process takes about 25 seconds.

25 seconds isn’t a big deal, right?

However, remember, I am doing this same thing 600 times!  Because I need to find the matching object FOR EACH of my 600 objects that I am adding discovery data for.

(600 * 25 seconds) = 15,000 seconds.  15000 seconds / 3600 seconds = 4.17 HOURS!!!!  This is no bueno!  I need to run this on a regular basis, and I cannot have it hammering my management servers like this.

 

So I take a different approach – what if I just go back to the SDK, and look up the Windows Computer object each time?

$MatchingComputerObject = Get-SCOMClass -Name "Microsoft.Windows.Computer" | Get-SCOMClassInstance | Where-Object -Property DisplayName -eq $NameImLookingFor

That takes 70 seconds and hits the SDK service and the SQL database every time.  Much worse!  Now my script will take 12 hours and hammer the SDK while it is running.  Ugh.

 

Enter in, the Hash Table

 

A Hash Table is defined as an associative array, with a compact data structure that stores key/value pairs.  Here is a simple example:

Create a new empty hashtable:

$HashTable = @{}

Add stuff to it:

$HashTable.Add('Kevin','Rocks')

Kevin is the “Key” in the Hashtable.  Rocks is the value.  In this case it is a simple string, but it could also be an object, which is powerful!

If I want to retrieve the Value for the Key “Kevin” I would:

$HashTable.'Kevin'

Which returns the value, “Rocks”.

Similarly – we can input an object:

$HashTable.Add('DC1.opsmgr.net',(Get-SCOMClass -Name "Microsoft.Windows.Computer" | Get-SCOMClassInstance | Where-Object -Property DisplayName -eq "DC1.opsmgr.net"))

In the above example, I am using the string FQDN as the Key, and inputting the Windows Computer object as the Value for the hashtable.

Now I can quickly retrieve the object without needing to go to the SDK, or loop through an array to find that object.

 

Here is a real world example:

# Get all the Windows Computers into my Array $WCarr = Get-SCOMClass -Name "Microsoft.Windows.Computer" | Get-SCOMClassInstance # Create an empty Hashtable $HashTable = @{} # Fill the hashtable from the array using the unique FQDN as the Key FOREACH($WC in $WCarr) { $HashTable.Add("$($WC.DisplayName)",$WC) } # Loop through each instance of my custom array and get properties from the Windows Computer as needed FOREACH ($MyObject in $MyObjects) { # Retrieve the individual matching object from the Hashtable $WC = $HashTable.($MyObject.DisplayName) # Get some property from the Windows Computer Object $IPAddress = $WC.'[Microsoft.Windows.Computer].IPAddress'.Value }

 

Now – when I need to retrieve a matching object – I don’t need to parse through all 10,000 objects each time, I simply request the specific object by the indexed Key from the hashtable.

 

How long does this take?

Well I am glad you asked!

The following line from above: 

$WC = $HashTable.($MyObject.DisplayName)

takes exactly 0.4 MILLISECONDS.

That’s right.  Milliseconds.

What took 21 SECONDS before, now takes four tenths of a millisecond.

The script section, which took 4.17 HOURS before, now completes in 0.24 SECONDS.

 

Script time matters.  Make sure yours aren't running any longer than they have to.

I really do recommend placing a start time and end time, and total runtime in all your SCOM scripts – to help you monitor these kinds of things closely:

https://blogs.technet.microsoft.com/kevinholman/2017/09/10/demo-scom-script-template/

Azure IP Advantage の対象範囲が Azure Stack に拡大

$
0
0

執筆者: Natalia Mackevicius (Director, Azure Stack)

このポストは、2018 年 1 15 日に投稿された Announcing the extension of Azure IP Advantage to Azure Stack の翻訳です。

 

Azure IP Advantage の対象範囲がAzure Stack にデプロイされたワークロードにも拡大されることになりました。Azure Stack を使用してハイブリッド クラウド環境を構築し、自社データセンターやホスト環境で Azure を利用されるお客様の場合、デジタル環境ならではのあまり馴染みのない知的財産権のリスクに直面することが出てきますが、Azure IP Advantage では、こうした Azure Stack でハイブリッド クラウドを利用するお客様に対し、オープン ソース ソフトウェアを含む Azure サービスへの無制限の補償と、10,000 件の特許に関する防衛ポートフォリオを提供します。

Azure Stack は、オンプレミス環境や非接続環境からクラウド サービスにアクセスするために利用するサービスです。たとえば、石油/ガス大手企業の Schlumberger (英語) は、掘削作業の改善に Azure Stack を活用しているほか、Saxo Bank では、Azure リージョンのない地域でも同じアプリケーション コードを再利用して主権や規則の問題に対応できるようにするために Azure Stack を使用しています。Azure Stack を使用すると、ハイブリッド クラウド環境で一貫したサービスや API のセットを使用してアプリケーションを実行できます。そして今回、その Azure IP Advantage による知的財産権保護のメリットをハイブリッド クラウド環境のお客様にもご利用いただけるようになりました。

Azure IP Advantage は、Azure Stack サービスに対してマイクロソフトが無制限の補償 (英語) を提供するものです。この対象には Azure サービスをサポートするオープン ソース ソフトウェアも含まれます。また、10,000 件のマイクロソフトの特許 (英語) の防衛ポートフォリオの利用も可能になり、Azure Stack の SaaS アプリケーションを保護できます。この防御ポートフォリオは、世界のクラウド特許ポートフォリオのトップ 3 (英語) に選ばれています。また、通常そのようなことはありませんが、マイクロソフトが特許を特許不実施主体 (NPE) に譲渡した場合は、お客様にスプリンギング ライセンスが無料で提供され、特許を保護できます。

クラウドは基幹業務アプリケーションで使用されることが多いため、お客様はクラウド ベンダーを選定する際にさまざまなことを考慮しなければなりません。Azure および Azure Stack をご利用いただくことで、お客様のハイブリッド クラウド ワークロードには、業界最高レベルの知的財産権保護プログラムである Azure IP Advantage が自動的に適用されるようになります。

 

マイクロソフト、法執行機関などとの連携により Gamarue (Andromeda) を撲滅

$
0
0

本記事は、Microsoft Secure ブログ “Microsoft teams up with law enforcement and other partners to disrupt Gamarue (Andromeda) (2017 12 4 日 米国時間公開) を翻訳したものです。


このたび、マイクロソフトのデジタル犯罪対策部門 (Digital Crimes Unit、DCU) は、マイクロソフトのセキュリティ研究者や世界各国の法執行機関 (英語) と連携し、Gamarue (英語) を撲滅したことを発表しました。Gamarue は、感染したコンピューターのネットワークで広く使用されてきたマルウェアで、Andromeda ボットネットと総称されます。

今回の成功は、2015 年 12 月にマイクロソフトの Windows Defender 研究チームと DCU が Gamarue 撲滅に向けて開始した Coordinated Malware Eradication (CME) キャンペーンが実を結んだものです。マイクロソフトは、インターネット セキュリティ企業 ESET (英語) と協力し、Gamarue マルウェアとそのインフラストラクチャを詳しく調査してきました。

44,000 以上のマルウェアのサンプルを分析した結果、無秩序に拡大していく Gamarue のインフラストラクチャが解明され、世界各国の法執行機関に、以下の詳細情報が提供されました。

  • ボットネットのコマンド アンド コントロール サーバーが使用する 1,214 個のドメインと IP アドレス
  • 464 種類のボットネット
  • 80 種類以上の関連マルウェア ファミリ

世界規模での連携により、ボットネットのサーバーがテイクダウンされ、世界最大級のマルウェアであった Gamarue は撲滅されました。2011 年以降 Gamarue が蔓延させた脅威には、次のようなものがあります。

世界各地のマルウェアの動向

過去 6 年の間、非常に活発なマルウェアだった Gamarue は、撲滅される直前までその勢いを落とすことはありませんでした。直近 6 か月間の Windows Defender のテレメトリからも、Gamarue が世界中に拡散していたことがわかります。

図 1. 2017 年 5 月から 11 月の世界各地における Gamarue の拡散状況

世界中に拡散している中で、Gamarue の遭遇回数の上位 10 か国は、その多くをアジア諸国が占めています。

図 2. 2017 年 5 月から 11 月の Gamarue 遭遇回数上位 10 か国

過去 6 か月間で Gamarue が検出またはブロックされたコンピューターの数は、月平均で 1,095,457 台ほどにのぼります。

図 3. 2017 年 5 月から 11 月に、Gamarue に遭遇したコンピューター数、IP アドレス数、固有ファイル数 (LNK の検出を除く)

Gamarue ボット

Gamarue は、サイバー犯罪の闇市場で Andromeda ボットとして知られています。ボットとは、感染したコンピューターを攻撃者が操作できるようにするプログラムです。他のさまざまなボットと同様に、Gamarue は攻撃者向けの犯罪キットとして販売されています。

犯罪キットには、以下のコンポーネントが含まれています。

  • ボット ビルダー: コンピューターに感染するマルウェア バイナリをビルドします
  • コマンド アンド コントロール アプリケーション: 攻撃者がボットの管理と操作を行えるようにする PHP ベースのダッシュボード アプリケーションです
  • ドキュメント: Gamarue ボットネットの作成方法が記載されています

感染したコンピューターのネットワークであるボットネットは、コマンド アンド コントロール (C&C) サーバーと呼ばれるコンピューター サーバーと通信します。攻撃者はこのサーバーを使用して、感染したコンピューターを操作します。

進化を続ける Gamarue に対し、セキュリティ研究者たちによって綿密な分析が行われてきました。テイクダウン直前には、活動中の Gamarue は 2.06、2.07、2.08、2.09、2.10 の 5 つのバージョンが確認されており、最新の 2.10 が最も活発でした。

Gamarue はモジュール式のため、犯罪キットに含まれるプラグインや別売りのプラグインによって機能を拡張できます。Gamarue のプラグインには、以下のようなものがあります。

  • Keylogger (150 ドル) – キー入力やマウス操作を記録し、ユーザー名やパスワード、金融情報などを盗み出します
  • Rootkit (犯罪キットに付属) – 攻撃目標のコンピューターで実行されているあらゆるプロセスにルートキット コードを挿入し、Gamarue の持続性を確保します
  • Socks4/5 (犯罪キットに付属) – インターネット上のコンピューターにマルウェアや指示を送るために、特定のコンピューターをプロキシ サーバー化します
  • Formgrabber (250 ドル) – Web ブラウザー (Chrome、Firefox、Internet Explorer) から送信されるデータを取得します
  • Teamviewer (250 ドル) – 攻撃者が感染したコンピューターを遠隔操作したり、デスクトップを覗き見たり、ファイルを転送したりできるようにする機能を持ちます
  • Spreader – リムーバブル ドライブ (USB 接続のポータブル HDD やフラッシュ メモリなど) を通じて Gamarue マルウェアを拡散する機能を追加します。また、更新プログラムをダウンロードしたサーバーでドメイン名生成 (DGA) 機能を使用します

Gamarue 攻撃のキル チェーン

Gamarue は長年にわたり、以下のようなさまざまな攻撃手法で拡散されていました。

  • リムーバブル ドライブ
  • Gamarue をホストする Web サイトへのリンクを含むソーシャル メディア (Facebook など) のメッセージ
  • ドライブバイ ダウンロード/エクスプロイト キット
  • 悪意のあるリンクを含むスパム メール
  • トロイの木馬型ダウンローダー

Gamarue に感染したマシンは C&C サーバーと通信し、ボットネットの一部となります。攻撃者は C&C サーバーを通じて感染したマシンを操作したり、情報を盗み出したり、コマンドを実行して他のマルウェア モジュールをダウンロードしたりします。

図 4. Gamarue 攻撃のキル チェーン

Gamarue の主な目的は、他の関連マルウェア ファミリを拡散させることです。CME キャンペーンの展開中は、少なくとも 80 種類のマルウェア ファミリが Gamarue によって拡散されました。このマルウェア ファミリには以下のようなものがあります。

他のマルウェアをインストールすることにより、感染したマシンのネットワークで攻撃者が活動できる範囲を拡大します。

C&C サーバーとの通信

Gamarue マルウェアに感染したマシンから C&C サーバーへの通信がトリガーされると、ハード ディスクのボリューム シリアル番号 (コンピューターのボット ID として使用)、Gamarue のビルド ID、感染したマシンのオペレーティング システム、ローカル IP アドレス、サインインしたユーザーが管理者権限を保持しているかどうかを示すインジケーター、感染したマシンのキーボードの言語設定などの情報が収集されます。この情報は、JSON 形式で HTTP を経由して C&C サーバーに送信されます。

図 5. Gamarue から C&C サーバーに送信される情報

この中で興味深いのは、キーボードの言語設定に関する情報です。以下の国に対応するキーボード言語が設定されているコンピューターは、それ以上感染が拡大することはありません。

  • ベラルーシ
  • ロシア
  • ウクライナ
  • カザフスタン

これらの情報は、C&C サーバーに送信する前に Gamarue マルウェア本体にハードコードされているキーを使用して RC4 アルゴリズムで暗号化されます。

図 6. C&C サーバーとの暗号化された通信

C&C サーバーがメッセージを受信すると、攻撃者が制御用ダッシュボードで事前に割り当てたコマンドを送信します。

図 7. 攻撃者が Gamarue ボットとの通信に使用する制御用ダッシュボードのサンプル

送信されるコマンドは、以下のいずれかです。

  • Download EXE (EXE ファイルのダウンロード。追加の実行可能なマルウェア ファイルなど)
  • Download DLL (DLL のダウンロード。追加のマルウェアなど。バージョン 2.09 以降では削除された)
  • Install plugin (プラグインのインストール)
  • Update bot (ボットの更新。ボット マルウェアなど)
  • Delete DLLs (DLL の削除。バージョン 2.09 以降では削除された)
  • Delete plugins (プラグインの削除)
  • Kill bot (ボットの停止)

下 3 つのコマンドは、マシンから Gamarue の痕跡を消すために使用されます。

C&C サーバーからの返信も、感染したマシンからのメッセージの暗号化と同じキーを使用して、RC4 アルゴリズムで暗号化されます。

図 8. 暗号化された C&C サーバーからの返信

復号化された返信には以下の情報が含まれています。

  • 時間間隔 (分単位) – C&C サーバーに次のコマンドを要求するまでの待機時間
  • タスク ID - タスク実行時のエラーの有無を追跡するために攻撃者が使用する ID
  • コマンド – 上記のコマンドのいずれか
  • ダウンロード URL - コマンドに応じてプラグイン、更新されたバイナリ、他のマルウェアをダウンロードする URL

図 9. 復号化された C&C サーバーからの返信

サンドボックス対抗技術

Gamarue では、分析や検出を困難にするためにウイルス対策ツールに対抗する手法が用いられています。感染する前に、対象マシンで実行中のプロセスのハッシュ リストをチェックして、仮想マシンやサンドボックス ツールなどのマルウェア分析ツールに関連するプロセスを発見した場合は、そのマシンに感染するのを止めます。古いバージョンでは、仮想マシンの場合には偽のペイロードが出現するようになっています。

図 10. Gamarue がマルウェア分析ツールに関連したプロセスが実行されていないかをチェック

ステルス技術

Gamarue では、検出を回避するためにクロスプロセス インジェクション (英語) 技術が使用されており、以下のようなコードを正規のプロセスに挿入します。

  • msiexec.exe (Gamarue バージョン 2.07 ~ 2.10)
  • wuauclt.exe、wupgrade.exe、svchost.exe (バージョン 2.06)

また、ルートキット プラグインを使用して Gamarue のファイルと Autostart レジストリ エントリを潜ませます。

さらに、プラグインの保存と読み込みにもステルス技術を使用します。プラグインはレジストリまたは Gamarue のファイルの代替データ ストリームのいずれかにファイルレスで保存されます。

OS の改ざん

Gamarue は、感染したコンピューターのファイアウォールや Windows Update、ユーザー アカウント制御機能を無効化し、オペレーティング システムを改ざんしようとします。これらの機能は、Gamarue がマシンから駆除されるまで再度有効化することができません。Windows 10 であれば、OS の改ざんを行うことはできません。

図 11. 無効化されたファイアウォールと Windows Update

収益化

攻撃者が Gamarue で金銭を得る方法は複数あります。関連マルウェアを拡散させることが主な目的である Gamarue は、1 回のインストールごとに金銭が支払われるスキームを利用して利益を上げます。また、プラグインを使用してユーザー情報を盗み出し、サイバー犯罪の闇市場で他の攻撃者に販売します。Gamarue に感染したコンピューターへのアクセス自体も、犯罪者グループ間で販売、貸し出し、リース、交換が行われます。

修復

Gamarue およびその他のマルウェアや望ましくないソフトウェアの感染を防ぐには、以下のような対策を実施します。

  • 未知のユーザーからのメールやソーシャル メディアのメッセージの開封に注意する
  • プログラム開発業者以外の Web サイトからソフトウェアをダウンロードする場合は慎重に行う

最も重要なのは、Gamarue などの脅威からコンピューターを保護する適切なセキュリティ ソリューションを使用することです。Windows Defender ウイルス対策は、Gamarue マルウェアを検出し削除します。Windows Defender ウイルス対策では、高度な機械学習モデル、ジェネリックおよびヒューリスティックな技術を使用して、新規および未知のマルウェア (英語) をクラウド保護サービスでリアルタイムに検出します。このほか、Microsoft Safety ScannerMalicious Software Removal Tool (MSRT) などの単体ツールでも Gamarue を検出して削除することができます。

Microsoft Edge は、ソーシャル メディアのメッセージに含まれる悪意のあるリンク、ドライブバイ ダウンロード、エクスプロイト キットなどによる Web サイトからの Gamarue 感染を防ぎます。保護されたブラウザーである Microsoft Edge は、権限の低いアプリ コンテナー内でページを開き、評価に基づいて悪意のあるダウンロードをブロックすることができます。

企業向けの環境では、保護をさらに強化することができます。Windows Defender Advanced Threat Protection では、セキュリティ運用担当者がクロスプロセス インジェクションなどの Gamarue のアクティビティをネットワークで検出し、攻撃を調査および対処する作業を支援します。行動分析や機械学習によって強化された検出ライブラリを備えた Windows Defender ATP では、配信やインストールから、永続化メカニズム、C&C サーバーとの通信に至るまで、マルウェア感染プロセス全体で悪意のある行動を検出することができます。

Microsoft Exchange Online Protection (EOP) では、Office 365 ユーザーを保護する組み込み済みのスパム対策フィルタリング機能により、メールからの Gamarue 感染を防ぎます。Office 365 Advanced Threat Protection は、クリック時の保護機能により、危険な添付ファイル、悪意のあるリンク、ファイルへのリンクを含むメールからメールボックスを保護します。

Windows Defender Exploit Guard では、悪意のあるドキュメント (Gamarue 拡散するものなど) やスクリプトをブロックすることができます。Windows Defender Exploit Guard の攻撃表面縮小 (ASR) 機能では、組み込み済みのインテリジェンスのセットで、悪意のあるドキュメントで観察された不審な行動をブロックします。ASR ルールは、悪意のある添付ファイルをブロックして Microsoft Outlook や Web メール (Gmail、Hotmail、Yahoo など) からの実行や起動を阻止することもできます。

マイクロソフトは、Gamarue に感染したコンピューターの駆除に取り組んでおり、Virus Information Alliance (英語) でサンプルを含むワンタイム パッケージを提供するなど、企業の顧客保護を継続的に支援しています。

 

マイクロソフト デジタル犯罪対策部門および Windows Defender 研究チーム

 

Gamarue (Andromeda) 撲滅に関しては、以下の情報もご参照ください。

 

 

クラスターの検証を実行した際に、Hyper-V 統合サービスのバージョン検証で警告が発生する

$
0
0

こんにちは。Windows プラットフォーム サポートです。
Windows Server 2016 Hyper-V クラスター環境でクラスターの検証を実行した際に、Hyper-V 統合サービスのバージョン検証で以下の警告が発生することがあります。

-----------------------------------------

Hyper-V 統合サービスのバージョンの検証

結果 : 警告
警告 : 次の仮想マシンはホスト コンピューターと一致しない統合サービスを実行しています。この仮想マシンまたはホスト コンピューターの統合サービスを更新して、同じバージョンにする必要があります。

-----------------------------------------

[原因]
Windows Server 2012 R2 Hyper-V 以前の環境は、ゲスト OS の統合サービスとホスト OS の統合サービスのバージョンと一致させることを推奨しております。
(ホスト OS の統合サービスバージョンが更新された場合に、ゲスト OS の統合サービスバージョンを更新させることを推奨しております)

そのため、Hyper-V クラスター環境では、クラスターの "クラスターの検証" にて仮想マシンの統合サービスのバージョンを検証しており、
ホスト OS とゲスト OS の統合サービスのバージョンが一致しているかを確認しております。

しかしながら、Windows Server 2016 Hyper-V 環境では、Windows Server 2012 R2 Hyper-V 以前の環境とは異なり、ゲスト OS の統合サービスはホスト OS の統合サービスのバージョンとの一致させる必要はなくWindows Update により最新を適用することを推奨しております。
(ホスト OS 側の統合サービスのバージョンは廃止され、統合サービスのバージョンは "0.0" となります)

そのため、"クラスターの検証" で行う仮想マシンの統合サービスのバージョン確認を行うと、Windows Server 2016 Hyper-V では
ホスト OS の統合サービスのバージョンが 0.0 のため、常にゲスト OS の統合サービスのバージョンと不一致となり、上記の警告が出力されます。

[対処について]
クラスターの検証で上記の警告が発生しても、クラスターの操作や動作等に影響を与えないことから無視してください。

本 Blog が少しでも皆様のお役に立てれば幸いです。

Robocopy のエラー (戻り値) について

$
0
0

皆さん、こんにちは。Windows プラットフォーム サポートの吉田です。

robocopy コマンドでファイル コピーを実行した際、LOG オプションで出力されるログに戻り値が記録されますが、今回は戻り値の結果の詳細についてご紹介します。

まず、robocopy コマンドでは以下のベースとなる戻り値があります。

戻り値 0: コピーする必要がないため、何も実施しなかった
戻り値 1: ファイルのコピーが成功した (フォルダーのコピーは含まれません)
戻り値 2: 余分なフォルダー、ファイルが確認された (コピー元にはなく、コピー先だけにある)
戻り値 4: 同じ名前で別の種類のファイルが存在した (コピー元はフォルダーで、コピー先はファイル、またはその逆)
戻り値 8: コピーに失敗した (リトライした結果を含みます、また /L では実際にコピー処理を行わないため、実質 8 以上の戻り値は出力されません)

このそれぞれの戻り値は LOG オプションでカウントされる場所は以下となります。

0 と判定されたファイル、フォルダーはログ中の "スキップ" にカウントされます。
1 と判定されたファイル、フォルダーはログ中の "コピー済み" にカウントされます。
2 と判定されたファイル、フォルダーはログ中の "Extras" にカウントされます。
4 と判定されたファイル、フォルダーはログ中の "不一致" にカウントされます。
8 と判定されたファイル、フォルダーはログ中の "失敗" にカウントされます。

しかし、robocopy の Log ファイルをご覧いただいた事のある方であれば、戻り値が 9 以上となっている結果をご覧いただいたことがあるかもしれません。
これは、複数のファイル、フォルダーをコピーされる場合に別々の結果となった場合に、足し算された値が返されるためです。

例えば、戻り値が [1] の場合には、robocopy によって処理されたファイルが、すべて正常に "コピー済み" と判断された場合です。
戻り値が [3] の場合、robocopy によって処理されたファイルの中に、戻り値 1 である "コピー済み" と戻り値 2 である "Extras" と判断されたファイルが混在する場合に記録されます。
つまり、戻り値が [1] 以外となっている場合には、正常にコピーが完了しなかったファイルが存在していることを表します。

以下に、[1] 以外のそれぞれの戻り値の結果をご紹介いたしますので、ご参考願います。

戻り値 3: 一部のファイルのコピーに成功したが、一部、Extras と判定された。(1 + 2)
戻り値 5: 一部のファイルのコピーに成功したが、一部、不一致 と判定された。(1 + 4)
戻り値 6: ファイルのコピーに成功しておらず、Extras または 不一致 と判定された (2 + 4)
戻り値 7: 一部のファイルのコピーに成功したが、一部 Extras または 不一致 と判定された (1 + 2 + 4)
戻り値 9: 一部のファイルのコピーに成功したが、一部 失敗 と判定された (1 + 8)
戻り値 10: ファイルのコピーに成功しておらず、Extras または 失敗 と判定された (2 + 8)
戻り値 11: 一部のファイルのコピーに成功したが、一部 Extras または 失敗 と判定された (1 + 2 + 8)
戻り値 12: ファイルのコピーに成功しておらず、不一致 または 失敗 と判定された (4 + 8)
戻り値 13: 一部のファイルのコピーに成功したが、一部 不一致 または 失敗 と判定された (1 + 4 + 8)
戻り値 14: ファイルのコピーに成功しておらず、Extras、不一致 または 失敗 と判定された (2 + 4 + 8)
戻り値 15: 一部のファイルのコピーに成功したが、一部 Extras、不一致 または 失敗 と判定された (1 + 2 + 4 + 8)

戻り値 16: ヘルプを表示したときにセットされます。また、存在しないフォルダーなどを指定するなど、引数が不正な場合にも記録されます。

このように、戻り値での結果では、すべてのコピー対象についてどのような結果が判定されたかを全体で把握するのみとなりますため、詳細な結果は LOG オプションによる出力結果を確認する必要がございますのでご留意願います。

以下は参考となりますが、/LOG オプションによる出力結果例となります。

-------------------------------------------------------------------------------
ROBOCOPY     ::     Windows
の堅牢性の高いファイル コピー
-------------------------------------------------------------------------------

  開始: Mon Jan 22 22:14:52 2018

   コピー元 : \dcshare
コピー先 : c:dest

    ファイル: temp*.txt

オプション: /TS /FP /BYTES /S /E /COPY:DAT /DCOPY:T /PURGE /NP /IT /MT:4 /R:6 /W:10 

------------------------------------------------------------------------------

                       2    \dcshare
新しいファイル                 0 2018/01/22 07:19:05    \dcsharetemp01.txt
0%
2018/01/22 22:14:52
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。


新しいファイル                 0 2018/01/22 07:19:12    \dcsharetemp02.txt
100%  10
秒間待機しています...
新しいファイル                 0 2018/01/22 07:19:05    \dcsharetemp01.txt 再試行しています...
2018/01/22 22:15:02
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。

10 秒間待機しています... 再試行しています...
2018/01/22 22:15:12
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。

10 秒間待機しています... 再試行しています...
2018/01/22 22:15:22
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。

10 秒間待機しています... 再試行しています...
2018/01/22 22:15:33
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。

10 秒間待機しています... 再試行しています...
2018/01/22 22:15:43
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。

10 秒間待機しています... 再試行しています...
2018/01/22 22:15:53
エラー 32 (0x00000020) ファイルをコピーしています \dcsharetemp01.txt
プロセスはファイルにアクセスできません。別のプロセスが使用中です。


エラー: 再試行が制限回数を超えました。


------------------------------------------------------------------------------

                  合計     コピー済み      スキップ       不一致        失敗    Extras
ディレクトリ:         1         0         1         0         0         0
ファイル:         2         1         0         0         1         0
バイト:         0         0         0         0         0         0
時刻:   0:01:00   0:00:00                       0:00:00   0:01:00

       終了: Mon Jan 22 22:15:53 2018

 

[アドレス一覧に表示しない] と設定したユーザーや連絡先の情報をアドインから取得することができない現象について

$
0
0

こんにちは。日本マイクロソフト Outlook サポート チームです。

グローバル アドレス帳の情報で、表示させたくないユーザーやグループ、連絡先などについて [アドレス一覧に表示しない] と設定することにより、グローバル アドレス帳に表示させないように設定することができます。
しかし、このような設定を行っている場合、アドインやマクロで Outlook Object Model を利用して情報を取得する場合に SMTP アドレスなどの情報が取得できないことがあります。

一部のユーザーに特定のユーザーやグループ、連絡先などについて表示させたくないが、その他のユーザーでは表示を行いたい、かつアドインでは情報を取得したいという場合、アドレス帳 ポリシーなどでアドレス帳を分け、表示内容を変えることができます。

- 参考情報
シナリオ: アドレス帳ポリシーの展開

アドレス帳ポリシーの作成

メール ユーザーへのアドレス帳ポリシーの割り当て

例えば、User A についてアドレス帳 ポリシーの設定により User B にはアドレス帳で非表示としても、User C には表示が行われるよう設定することができます。
User C についてはアドインで User A の情報を取得することができます。

本情報の内容 (添付文書、リンク先などを含む) は作成日時点でのものであり、予告なく変更される場合があります。

Future Computed:AI とその社会における役割

$
0
0


[ブログ更新日:2018年1月17日]

ブラッドスミス (Brad Smith) 
プレジデント兼最高法務責任者
ハリー シャム (Harry Shum) 
Microsoft AI and Research Group 担当 エグゼクティブバイスプレジデント

マイクロソフトは新しい電子書籍「Future Computed:AIとその社会における役割」(“The Future Computed: Artificial Intelligence and its role in society”)を公表しました。私たちが前書きを担当し、私たちのチームが執筆に協力しています。タイトルが示すように、本書は、AIテクノロジの方向性、そして、それがもたらす社会的課題についてのマイクロソフトの見解を述べたものです。

個人的レベルで言えば、前書きを担当したことで、過去20年間にテクノロジがどれほど私たちの生活を変えてきたか、そして今後20年間にどのような変化が起こるのかを、一歩離れて熟考する機会が得られました。1998年に私たちは2人ともマイクロソフトに勤務していましたが、勤務地は地球の裏側でした。別の大陸のまったく異なる文化の中で暮らしていましたが、人の手による計画と活動が中心の生活という点では共通でした。それから20年経った今、かつてはSFと考えられていたデジタルの世界が当たり前の存在になっています。

モバイルデバイスやクラウドコンピューティングなどのテクノロジが私たちのニュースの読み方、日々の計画、コミュニケーション、ショッピング、そして、家族、友人、同僚とのやり取りを根本的に変革しました。今から20年後にはこの世界はどのようになっているでしょうか?マイクロソフトは、AIにより、私たちの最も貴重な資源、つまり「時間」を節約しながらより多くのことができるようになると考えています。2038年までには、パーソナルデジタルアシスタントが私たちのニーズを先読みし、スケジュールを管理し、会議の準備をし、日々の計画を助け、コミュニケーションを行い、自動車を運転してくれるようになるでしょう。

個人の生活以外でも、AIはヘルスケア、農業、教育、運輸などの領域でブレークスルーを実現していきます。既に驚くべき事例が存在しています。

しかし、過去20年間においてもそうであったように、新しいテクノロジは必然的に複雑な問題と広範な社会的懸念をもたらします。コンピューターと人間の協力によって生まれる未来を考える上では、これらの課題に正面から対応することが重要です。

AIが責任ある形で作られ、使用されるためにはどうすればよいでしょうか?人々を守るための倫理基準をどのように確立すればよいのでしょうか?AIの利用基準をどう統制すべきでしょうか?そして、AIは雇用と職業にどのような影響を与えるのでしょうか?

これらの難しい質問に答えるために、科学技術者は政府、学術界、企業、市民社会などの利害関係者との連携を強めていく必要があります。マイクロソフトでは、学際的な人工知能の開発と利用の指針として公平性、信頼性、プライバシーと安全性、協調性、透明性、そして、説明責任という6つの倫理基準を定めています。これらの課題をより深く理解するほど、そして、テクノロジの開発者によるこれらの課題対応のベストプラクティスの共有が進むほど、AI統制の社会的ルールが熟考され、世界に対する貢献度は高まるでしょう。

また、AIの労働市場への影響にも注意を払う必要があります。AIによって不要になる職業は何でしょうか?AIはどのような職業を生み出すのでしょうか?250年にわたるテクノロジの変化の中で変わらない要素があるとするならば、それはテクノロジの労働市場への絶え間ない影響、すなわち、新しい職種の誕生、既存職種の消滅、そして、職業の特性の変化です。この影響が今後も続いていくのは確実です。

今、いくつかの重要な事実が明らかになりつつあります。

第一に、AI時代には、AIによる変化を迅速かつ効果的に取り入れた企業や国家が成長するということです。なぜなら、新しい職業や経済成長はテクノロジを採用する者に訪れるからです。

第二に、AIが重要な社会的問題の解決に貢献できる一方で、その未来についてはより批判的な目で見なければならないということです。機会と同様に課題もあります。強固な倫理基準、法律の改正、新たなスキルの訓練、さらには、労働市場の改革などに対応しなければなりません。AIから最大の価値を得るためには、これらの課題すべてに対応する必要があります。

第三に、AIはテクノロジセクターだけで作られるものではないことから、共同責任の意識を持つことが必要とされるということです。マイクロソフトでは、かつてPCをすべての人に提供したのと同様に、AIを民主化しようとしています。つまり、開発者、企業、政府が容易にAIベースのソリューションを構築し、社会に迅速に価値を提供できるようにするためのツールを作成しています。

これらの点は最も重要な結論へと結びつきます。それは、AIによる世界のためのスキルの獲得には、科学、テクノロジ、工学、数学以外の要素も必要ということです。コンピューターがより人間に近くなるにつれ、社会科学や人類学が今まで以上に重要になるでしょう。言語学、文芸、歴史、経済学、倫理、哲学、心理学、人格形成などの科目によって、AIソリューションの開発と管理で重要になる哲学的・倫理的なスキルを教育することができます。AIが人類に最大限貢献できるようにするためには、すべてのエンジニアが一般教養をより深く学び、すべての一般教養学部の学生がより深く工学を学ぶことが必要になります。

私たちには先見の明があるわけではありませんが、私たちが互いの言葉を聞き、学ぶためにより多くの時間を費やさなければならないことは確かです。「計算される未来」がこの対話に貢献できることを願っています。

「Future Computed」はこちらから入手可能です。また、追加の情報はこちらから参照可能です。

ーーー

本ページのすべての内容は、作成日時点でのものであり、予告なく変更される場合があります。正式な社内承認や各社との契約締結が必要な場合は、それまでは確定されるものではありません。また、様々な事由・背景により、一部または全部が変更、キャンセル、実現困難となる場合があります。予めご了承下さい。


WSUS クライアントのグループ ポリシー : その 3 – Windows 10 基本編

$
0
0

こんにちは。WSUS サポートチームです。

今回は、Windows 10 WSUS クライアントとして管理する場合のオススメのグループ ポリシーの設定や設定時の注意点を、下記のシナリオ毎に紹介します。Windows 10 の導入を検討されている方や既に導入されている方は是非ご一読ください!

 

A. 更新プログラムを WSUS サーバーのみから取得するようにしたい場合

B. 更新プログラムの自動適用は WSUS サーバーから行なうが、外部へ端末を持ち出した際等に手動で Windows Update / Microsoft Update サイトより更新プログラムの取得を行えるようにしたい場合

 

A. 更新プログラムを WSUS サーバーのみから取得するようにしたい場合

Windows 10 のクライアントにて、WSUS サーバーからのみ更新プログラムを取得させたい場合には、下記の 3 種類のグループ ポリシーを有効に設定してください。

 

1. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [イントラネットの Microsoft 更新サービスの場所を指定する]

2. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [自動更新を構成する]

3-1. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [インターネット上の Windows Update に接続しない]

もしくは

3-2. [コンピューターの構成] -> [管理用テンプレート] -> [システム] -> [インターネット通信の管理] -> [インターネット通信の設定]
-> [Windows Update のすべての機能へのアクセスをオフにする]

 

また、Window 10 バージョン 1607 の環境では、上記の設定に加え、下記の設定を有効にすることをご検討ください。

 

4. デュアル スキャン対策のため KB4025334  (OS Build 14393.1532) 以降の更新プログラムを適用後

4-1. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [Windows Update に対するスキャンを発生させる更新遅延ポリシーを許可しない]

および

4-2. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update] -> [Windows Update   延期]
    -> [機能更新プログラムをいつ受信するかを選択してください]
※ ブランチ準備レベルや延期日数については任意の値で構いません。

 

1、2 のグループ ポリシーについては、こちらの記事で詳細を紹介しているため、本記事では 3、4 について詳細を紹介していきます。

 

< 3. [インターネット上の Windows Update に接続しない] もしくは [Windows Update のすべての機能へのアクセスをオフにする] >

このいずれかのグループ ポリシー設定しないと、設定画面上にてユーザーが以下の [オンラインで Microsoft Update から更新プログラムを確認します。] ボタンを押した場合には、WSUS サーバーではなく Windows Update / Microsoft Update サイトより更新プログラムを取得してしまいます。このため、WSUS サーバーのみから更新プログラムを取得させたい場合には、必ずいずれかの設定をしていただく必要があります。


また、[インターネット上の Windows Update に接続しない] を有効にした場合は、ストアアプリのインストールや更新ができません。ストアアプリのインストールや更新を許可したい場合は、[Windows Update のすべての機能へのアクセスをオフにする] の方を有効にしてください。

 

< 4. デュアル スキャン対策の設定 >

Windows 10 バージョン 1607 の環境においては、上記の 3 種類の設定を行っても、ユーザーが下記のチェックボックスにチェックを入れた場合には、WSUS からではなく Windows Update / Microsoft Update サイトより更新プログラムを取得してしまいます。これはこの記事で詳細を紹介している通り、下記のチェックを有効化することで、Windows Update for Business が有効化され、デュアル スキャンが発生するためです。

上述の 4 の設定を行うことで、上記のチェック ボックスがグレー アウトするため、ユーザー操作によってデュアル スキャンが動作することを防ぐことが出来ます。

なお、Windows 10 バージョン 1703 以降の環境では、この記事の「補足3:UIの表示について」で紹介している通り、1 の設定を行うと上記の設定に対応する項目は非表示となるため、本設定をしていただく必要はありません

 

B. 更新プログラムの自動適用は WSUS サーバーから行なうが、外部へ端末を持ち出した際等に手動で Windows Update / Microsoft Updateサイトより更新プログラムの取得を行えるようにしたい場合

基本的には WSUS サーバーから更新プログラムの自動配信が行われるようにしつつも、外部へ端末を持ち出したタイミング等で WSUS サーバーと通信が行えない環境では、ユーザー操作によって Windows Update / Microsoft Update サイトより更新プログラムを取得出来るようにしたいという場合には、以下の設定を行ないます。

 

1. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [イントラネットの Microsoft 更新サービスの場所を指定する]

2. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [自動更新を構成する]

3. デュアル スキャン対策のため Windows 10 1607 の場合、KB4025334  (OS Build 14393.1532) 以降の更新プログラム、Windows 10 1703 の場合、 KB4041676 (OS Build 15063.674) 以降の更新プログラムを適用後

3-1. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [Windows Update に対するスキャンを発生させる更新遅延ポリシーを許可しない]

および

3-2. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update] -> [Windows Update の  延期 / Windows Update for Business]
    -> [機能更新プログラムをいつ受信するかを選択してください / プレビュー ビルドや機能更新プログラムをいつ受信するか選択してください]

 

こちらについても 1、2 のグループ ポリシーについては、この記事で詳細を紹介しているため、本記事では 3 について詳細を紹介していきます。

 

< 3. デュアル スキャン対策の設定 >

本シナリオの場合、WSUS サーバーからだけではなく Windows Update / Microsoft Update サイトからも更新プログラムを受信するため、下記のグループ ポリシーの設定を検討していただく必要があります。Windows Update / Microsoft Update サイトでは常に最新の更新プログラムが配信されているため、下記の設定をしないと、WSUS 上で配信の制限を掛けている更新プログラムやアップグレードが、ユーザーが [オンラインで Microsoft Update から更新プログラムを確認します。] をクリックしたタイミングで配信されてしまうことがあります。

 

3-2. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update] -> [Windows Update   延期 / Windows Update for Business]
    -> [機能更新プログラムをいつ受信するかを選択してください / プレビュー ビルドや機能更新プログラムをいつ受信するか選択してください] ※ 本設定についてはこの記事でも詳細を紹介しております。

 

また、上記の設定に加え下記の設定を行わないと、この記事で詳細を紹介している通り、Windows Update for Business が有効化され、デュアル スキャンの動作が発生してしまいます。

 

3-1. [コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
-> [Windows Update に対するスキャンを発生させる更新遅延ポリシーを許可しない]

 

上記の設定を全て行うことで、自動更新や下記の設定画面の [更新プログラムのチェック] を押したタイミングでは、WSUS 上で承認している更新プログラムが全て配信され、[オンラインで Microsoft Update から更新プログラムを確認します。] をクリックした場合には、3-2 で指定した延期日数やブランチの設定に合わせて機能更新プログラムの配信が行われます。

 

補足 : デュアル スキャンに関する注意事項

この記事で紹介している通り、下記の 3 つのグループ ポリシーは上述のデュアル スキャン対策と合わせて設定しないと、WSUS サーバーからではなく、Windows Update / Microsoft Update サイトから更新プログラムを取得する動作が発生してしまいますのでご注意ください。

 

[コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update] -> [Windows Update の延期]
-> [
機能更新プログラムをいつ受信するかを選択してください / プレビュー ビルドや機能更新プログラムをいつ受信するか選択してください]
-> [
品質更新プログラムをいつ受信するかを選択してください]

[コンピューターの構成] -> [管理用テンプレート] -> [Windows コンポーネント] -> [Windows Update]
->  [Windows Update
からドライバーを除外する]

 

Microsoft Premier Workshop: SQL Server 2017 für Developer

$
0
0

Der neue SQL Server 2017 bietet bahnbrechende Neuigkeiten und Verbesserungen für Applikations- und Datenbankenentwickler. Unter anderem auch zum ersten Mal das Graph Datenmodell, welches sich in die relationale Datenbank Engine integriert und mit bekannten SQL Befehlen sehr effizient abfragen lässt. Die 2017 Version bietet außerdem die Integration von Python und als Highlight die Verfügbarkeit des neuen SQL Server auf dem Betriebssystem Linux und in Docker Containers an.
Setzen Sie den neuen Server auf der Plattform an, die sie mögen. Nutzen Sie Entwicklungstools, die Sie schon kennen. Modellieren Sie die Daten so wie es Ihre Anwendungen brauchen, und bei jeder dieser Kombination unterstützt sie der SQL Server 2017 mit der gleichen leistungsstarken Datenbank Engine.

Erlernen sie in diesem Workshop wie sie diese und viele weitere neuen Features und Optionen effektiv nutzen können. Weiter Details zum Workshop finden sie auch im beigefügten Datenblatt.

Agenda

Modul 1: Introduction to SQL Server 2017
Überblick über die Neuerungen und Änderungen in SQL Server 2017.

Modul 2: Performance Improvements
Adaptive Query Processing und Automatic Tuning. In-Memory Performance und Operational Analytics, Columnstore Indexes, Query Store, Live Query Statistics und verbesserte JSON-Funktionen. Dazu werden die neuen Clusterless Read-Scale Availability Groups vorgestellt.

Modul 3: Database Server Programmability
Das neue Feature Graph Daten wird in diesem Modul vorgestellt. Neuerungen in der Programmiersprache Transact-SQL: Temporal Data, neue Funktionen und Resumable Online Index Rebuild.

Modul 4: Security Upgrades
Implementierung von Security Features: Always Encrypted, Row-Level Security und Dynamic Data Masking.

Modul 5: Application Development
Einführung in die Entwicklung von Datenbanken und Database Lifecycle Management mit SSDT bzw. Visual Studio 2017.

Modul 6: Enterprise Information Management
Verbesserungen in Integration Services: Scale-out Option, SSIS für Linux und die neue Azure Data Factory V2.

Modul 7: Business Intelligence
Neuerungen und Erweiterungen in den Reporting Services und Analysis Services: DAX Support, Power BI Integration, Object-Level Security, neue Datenquellen und Enhanced Ragged Hierarchy Support.

Modul 8: Machine Learning Services
Integration von Python und R in das SQL Server Ökosystem und die neue T-SQL Funktion Native Scoring. Machine Learning Services als umfangreiche Plattform für die Entwicklung moderner KI-Anwendungen.

Modul 9: SQL Server on Linux and Docker
Der neue SQL Server 2017 auf anderen Betriebssystemen. Installation, Konfiguration und Best Practices. Funktionalitäten, Unterschiede und Einsatzszenarien.

 

Zielgruppe
Dieser Workshop richtet sich an Architekten und Entwickler, die für die Planung und Implementierung von SQL Server Lösungen verantwortlich sind. Alle Teilnehmer sollen die Konzepte der Relationalen Datenbanken verstehen und Erfahrung mit SQL Server oder ähnlichen Datenbanksystemen mitbringen.

Level 300
(Level Skala: 100= Strategisch/ 200= technischer Überblick/ 300=tiefe Fachkenntnisse/ 400= technisches Expertenwissen)

 

Anmeldung
Zur Anmeldung wenden Sie sich bitte direkt an Ihren Microsoft Technical Account Manager oder besuchen Sie uns im Web auf Microsoft Premier Education. Dort finden Sie eine Gesamtübersicht aller offenen Workshops, für die Sie sich dort auch gleich anmelden können.

Microsoft Premier Workshop Service Fabric for Developers

$
0
0

Software Methodologien, Prozesse, Muster, Plattformen und Geräte entwickeln sich ständig weiter. Viele Anwendungen durchlaufen die monatliche 2-tier oder 3-tier Entwicklung.
Microservices ist ein Ansatz der Logik und Zustand kombiniert und unabhängig versioniert, dies bereitstellt und skaliert.
Azure Services Fabric ist in Azure als Dienst und Lokal als eigenständige Clusterkonfigurationen verfügbar. Anwendungen, die für Programmier-APIs erstellt wurden, können in Azure, private Cloud oder anderen Cloud Anwendungen ohne Codeänderungen bereitgestellt werden. Microsoft entwickelte die Service Fabric-Plattform und nutzt diese Plattform als Teil vieler interner Azure-Dienste.
Anstatt einer herkömmliche monatliche n-Tier-Anwendung zu entwerfen, sollten sie schnell kleinere unabhängige Softwarekomponenten mit microservices und Anwendungen, die auf der Plattform ausgeführt werden, einschließlich hoher Verfügbarkeit, Selbstheilung, Integritätsüberwachung, Skalierung und einfachen Programmiermodellen.
Dieser Workshop deckt unter anderem folgende Punkte ab:

Technical Highlights

• Understand the benefits and challenges with a Microservices approach
• The Actor Programming Model – Stateless and Stateful Actors
• The Services Programming API – Stateless and Stateful Services
• Create a Secure Service Fabric Cluster
• Stateful service partitioning and backup
• Service Fabric health monitoring capabilities
• Closer look at Service Fabric Platform

 

Agenda

Module 1: Introduction to Microservices and Design Patterns
Architectural Evolution to microservices, characteristics of microservices, principles of microservices, benefits and challenges.
Module 2: Introduction to Azure Service Fabric
Azure Service Fabric platform and major benefits, stateless and stateful actors, stateless and stateful services and capacity planning.
Module 3: Service Fabric Internals
Service synchronization, design benefits / trade offs of reliable actors vs. reliable services, serialization, failover and resource balancing.
Module 4: Service Fabric Communications
Gain an understanding of how services communicate with each other and how to use different communication listener types. Also learn about the pros/cons of the different communication listener types.
Module 5: Security
Learn how to setup a secure cluster using certificates and Azure Active Directory. We will also discuss node-to-node and application based security.
Module 6: Data Partitioning and Data Patterns
Working with data in stateful services, reliable collections, best practices for data partitioning.
Module 7: Service Fabric Health
Azure Service Fabric health monitoring and operational telemetry
Module 8: Azure Service Fabric with Containers
Gain an understanding of the basics of Windows Containers and how they can be deployed on to a Service Fabric cluster. Also learn about doing development and deployment with the assistance of Docker for Windows.

 

Zielgruppe
Der Workshop richtet sich an alle Softwareentwickler/- Architekten, Line of Business Application Developer, die mindestens 1 Jahr Erfahrung mit C# and .NET (oder ähnliche Vorkenntnisse) mitbringen oder Microsoft Certified Technology Specialisten (MCTS) bzw. Microsoft Certified Professional Developer (MCPD)

Level 300
(Level Skala: 100= Strategisch/ 200= technischer Überblick/ 300=tiefe Fachkenntnisse/ 400= technisches Expertenwissen)

Anmeldung
Zur Anmeldung wenden Sie sich bitte direkt an Ihren Microsoft Technical Account Manager oder besuchen Sie uns im Web auf Microsoft Premier Education. Dort finden Sie eine Gesamtübersicht aller offenen Workshops, für die Sie sich dort auch gleich anmelden können.

A Cloud for Global Good

$
0
0

Logbucheintrag 180124:

Cloud Computing ist eine unglaubliche Chance für Wirtschaft und Gesellschaft, denn es stellt Services rund um die Uhr und rund um den Globus bereit. Gleichzeitig aber ist Cloud Computing auch eine große Herausforderung für Wirtschaft und Gesellschaft, denn die weltweite Zugriffsmöglichkeit auf Anwendungen und Daten stellt auch neue Anforderungen an unsere Rechtsauffassung. Staatsgrenzen verschwinden in der Cloud.

Das genau ist der Kern des Rechtsstreits, in dem sich Microsoft seit rund einem halben Jahrzehnt mit der US-Regierung befindet. Im so genannten „New York Search Warrant Case“ verlangen US-Behörden im Zusammenhang mit einer Fahndung Zugriff auf personenbezogene Daten, die auf einem Azure-Server in Dublin liegen. Wir haben die Herausgabe der Daten verweigert.

Microsofts Chef-Justiziar Brad Smith hat unsere Position von Anfang an klar vertreten: Daten auf Servern außerhalb der USA unterliegen nicht den Zugriffsrechten der US-Behörden. Demgegenüber sind die US-Behörden der Auffassung, dass Unternehmen mit Hauptsitz oder Niederlassung in den USA vollständig US-amerikanischem Recht unterliegen – und damit auch ihre Installationen außerhalb der Vereinigten Staaten.

Inzwischen ist der Fall beim Obersten Gerichtshof der USA angekommen, der am 27. Februar zu einer nächsten Anhörung zum Thema zusammenkommt und bis zum Sommer eine Entscheidung treffen will. In diesen Tagen hat Microsoft Schützenhilfe durch den Bundesverband der Deutschen Industrie, die Industrie- und Handelskammern sowie zuletzt durch den Hightech-Verband Bitkom erhalten, der zusammen mit anderen europäischen Industrieverbänden ein so genanntes Amicus-Schreiben beim Supreme Court eingereicht hat. Darin nehmen die Europäer die im US-amerikanischen Recht verbriefte Möglichkeit in Anspruch, zu einem Gerichtsverfahren Stellung zu nehmen, an dem sie nicht unmittelbar beteiligt sind. Als „Amicus Court“, als Freund des Gerichts, warnen die Verbände vor einem tiefgreifenden Eingriff in europäisches Datenschutzrecht, wenn der Gerichtshof dem Verlangen der US-Behörden stattgeben sollte.

Bitkom-Hauptgeschäftsführer Bernhard Rohleder bringt es in einer Stellungnahme auf den Punkt. "Den Unternehmen droht ein unauflösbares Dilemma", sagt er. Folgten sie einer solchen Anordnung der US-Behörden, brechen sie europäisches Recht. "Widersetzen sie sich einer solchen Anordnung, brechen sie US-Recht."

Microsoft hat das Dilemma früh erkannt. Deshalb haben wir an zwei Standorten in Deutschland die Azure-Rechenzentren komplett in die treuhänderische Verwaltung der Deutschen Telekom gegeben. Microsoft-Mitarbeiter haben keinen Zugriff – weder zu den Servern, noch zu den Anwendungen oder ihren Daten. Aber diese doppelte Sicherheit hat ihren Preis. Für viele Kunden ist die Gefahr nicht konkret genug, um die zusätzlichen Kosten, die durch die doppelte Absicherung gegeben ist, zu tragen.

Deshalb haben CEO Satya Nadella und Chef-Justitiar Brad Smith Ende 2016 eine Denkschrift unter dem Titel „A Cloud for Global Good“ vorgestellt, die 78 Empfehlungen in 15 Politikfeldern unterbreitet, um die Cloud auch in Zukunft zu einem besonders sicheren Terrain zu machen. "Cloud Computing bietet die Antwort auf einige der größten Herausforderungen in der Welt", sagte Brad Smith damals. "Doch wie so oft bei technologischem Fortschritt ergeben sich auch viele Fragen für die Gesellschaft." Dafür brauche es "ein ausgewogenes Paket aus politischen und technologischen Lösungen". Das Papier ist gerade jetzt besonders lesenswert. Hier ist der Link zum Download.

Übrigens: In diesen Tagen ist Brad Smith erneut in Deutschland, um für sein neues Buch „The Future Computed“, das er zusammen mit Harry Shum geschrieben hat, zu werben. Erneut geht es darin um die Herausforderungen, denen die Gesellschaft durch neue Technologien gegenübersteht. Denn auch der Einsatz von Artificial Intelligence bedarf der ethischen, rechtlichen und gesellschaftspolitischen Einordnung. „Wir sind überzeugt, dass AI überall dort Vorteile bringt, wo Erkenntnisse (Intelligence) benötigt werden.“ Die Unternehmen und Staaten, die die Konsequenzen für sich zu nutzen und zu ordnen wissen, werden ein deutlich höheres Wohlstandsniveau erreichen. Nach „Cloud for the Global Good“ also auch „AI for the Global Good“.

Tech Summit 2017: Create a modern workplace with Microsoft 365

$
0
0

Microsoft Tech Summit has returned to Birmingham, opening to over 3,000 people at the NEC. The opening keynote preceded numerous sessions on Microsoft technologies, from Azure to Teams, across two days of learning. Kicking off the event was Stella Chernyak, Senior Director for Microsoft 365 and Office 365 Enterprise Audience Marketing, talking about Microsoft 365.

The keynote started by looking at trends in the workplace - where increasingly we're seeing that companies are looking at how they can be enabling their employees to do the best work they can, and how they can proactively foster that culture.

Microsoft has demonstrated how it's working to achieve this via Modern Workplace. However, how can other companies use Microsoft 365 to create their own modern workplaces?

What is a Modern Workplace?

While every company is unique and has their own story, we do hear some common themes from our customers. Customers are demanding help in enabling everyone in their company while they go through their digital transformation.

There's also a modernisation aspect. Millennials have different expectations when it comes to work, wanting a bigger sense of purpose and the ability to be more flexible. Working at any time, at any place and on any device is the new normal. But there's work that has to be done to make this secure and easy for your users.

Company expectations of workers is also changing; Everyone is expected to be creative because it's ingenuity that makes companies stand apart. Anyone, whether they're in customer service or working as a financial representative, they should be able to work together.

As we talk about these trends, we should be thinking about potential threats to security. On one hand you want to enable flexible working, but you also have to protect your company from complex, constantly evolving threats. Nowadays you really need to have intelligence about potential breaches, and you need to have the strength and breadth to prevent them from affecting your wider systems.

Microsoft 365 is our answer to this; a complete, intelligent solution including Office 365, Windows 10, and Enterprise Mobility + Security, that empowers everyone to be creative and work together, securely. To accomplish this, Microsoft 365 focuses on 4 areas.

Unlock Creativity

How can we use technology like AI to make every employee in the office work together? Applications like Outlook use intelligence to help users prioritise important messages. As it learns based on your own reading patterns, it'll surface important emails in the 'focused' inbox, and less important ones in the 'other' inbox. This is a basic but important use of AI, as it allows you to be much more productive and ensure that you're not missing important emails in a deluge of other, less important emails.

Outlook can also pull important information to the top of the email. For example, if you receive a travel itinerary in an email, Outlook can read it and know what information is important enough to highlight. It's displayed in a smart card, and it'll show destinations, flight times, travel duration and more.

It's not just about being more productive though - it's about using this intelligence to keep you safe. An example is that Outlook can look at suspicious emails and flag them as dangerous, even when at first glance it might appear to be an email from someone you know. It's an extra line of defence that can prevent threats from making their way into your business.

Whether it's memes or code snippets, Teams is the perfect for fostering collaboration in the workplace. This is because you're not limited to posting text and images - you can embed things like Power BI dashboards, streaming video and working files. If you're working in a company that has offices in multiple countries, you may find yourself in a position where you want to use a PowerPoint, but it's in a language you don't understand. All you need to do is use the 'Translate Slides' option to create a copy in another language.

There are also tools that'll help if you don't have a knack for designing PowerPoint slides. For example, by adding multiple images into a slide at the same time, PowerPoint will present you with design ideas that integrate these images with your current slide, automatically picking photo size, detecting focal points and colour matching. Add on top of that support for 3D models and AR functionality, this is just the start of how it can help foster creativity within your company.

Built for Teamwork

Email and calendar with Exchange, connecting to people, content and apps with SharePoint, and using voice, video and chat with Skype and Teams. This is just a selection of the numerous tools on Microsoft 365 that promote teamwork.

Central to our teamwork strategy is Teams, a digital hub for teamwork that brings together people, meetings, documents, conversations and obligations into one campus. As of today there are over 150 integrations that you can use to customise your team's experience. It's at the centre of our investment with Office 365.

We have modern Skype infrastructure for enterprise voice, video and meetings, and Teams as a hero client for Office 365 customers. On top of that there will also be a release of Skype for Business Server in late calendar year 2018.

Stella then spoke with James Boswell, Global Director of End User Computing at Centrica, about their digital transformation and adoption of Microsoft 365.

As Centrica is an energy and services company, digital transformation is key to the business. They needed to have the right data available to them at the right time, use Power BI to make smarter business decisions, and use Office 365 to make sure people can collaborate together. They needed to make sure they had the right tools to do their jobs.

This was the start of a massive Microsoft 365 transformation. They launched a SharePoint for 52,000 internal users around the world as a corporate intranet, moved everyone onto Skype and stored their documents in the cloud.

How are they moving so fast? Their board wanted to adopt the latest digital technology to have an advantage. They did this with great support from Microsoft and their suppliers. They also made sure the business itself was changing, making sure they had buy-in across the company, from HR to customer services, to ensure that it would be a smooth transition.

Some interesting trends have emerged this year thanks to that. People are now working from home because they have the right technology to do so. They have a 'Make a great idea count' program that allows anyone in the company to provide feedback to anything regarding the business, with 5,000 active users and massive productivity gains. Offices all around the world are using Yammer to chat with people on the other side of the world, something that hadn't been practical before.

But how do they manage keeping on top of the evergreen features of Office 365? Their team has changed, hiring a head of customer experience to make sure they're keeping on top of the tech, plus two full time digital evangelists ensuring that people on all levels know how to use the technology. They're creating a network of digital champions.

Integrated for Simplicity

The aim of Microsoft 365 is to make things a lot simper. Minimise TCO access deployment, management and servicing, have a broad support for PC, Mac, iOS and Android platforms, and more. Make it easier for your users.

One such technology that really changes this experience is Windows AutoPilot, a collection of technologies used to set up and pre-configure new devices, getting them ready for productive use. From the IT pros' perspective, the only interaction required from the end user is to connect to a network and to verify their credentials. Everything past that is automated.

The initial set-up of a device is something we're all familiar with. Windows AutoPilot makes sure that the end user only has to enter the bare minimum of information to get up and running. It recognises when the new device is connecting to a corporate network, asking for your password and downloading security policies for software like Windows Defender, BitLocker and Windows Hello. What used to take hours of set up is now incredibly quick.

There's also Upgrade Readiness, which provides powerful insights and recommendations about the computers, applications, and drivers in your organisation. You can view the video below for more information on how this works.

With Co-management, you can choose whether your policies are managed on premises in the configuration manager, or in the cloud on Intune. There's also an option to have a mix of both, so you can make sure you're making life easier for your end users.

Intelligent Security

We build security in all layers of Microsoft 365. We get information about billions of threats on products like Azure, Windows and Exchange, and we're constantly monitoring for new threats.

One of the challenges we have is getting users to classify and protect their documents. To make this easier, we can set up a rule to make particular words flag documents as confidential. For example, if we were to indicate that 'mark 8' is a classified phrase, and we then used it in a document, Word would now know that the document is confidential and will set usage rights accordingly. This prevents documents containing sensitive information from being passed to people that shouldn't see them.

Similar rules can be used to prevent vendors from downloading copies of documents that contain particular sensitive information. These documents could be read in the browser but not downloadable, depending on your organisation's own policies.

A common issue when it comes to making improvements to security is how you convince your higher ups to make important changes. This is tricky if you're the most technical person and need a way to show what you mean. The easiest way to do this is with a number.

We can now give security scores, which look at six areas in particular and scores them out of 100, providing recommendations for each of them should further action be required. With a score of 401 out of 600, it shows people who aren't as tech savvy that there are things that need to be improved.

Preparing for GDPR

Because of GDPR, there's an increased concern over data protection and privacy. We're working closely with both regulators and customers to make sure our products and services are within these regulations.

We want to help you focus on your core business while efficiently preparing for the GDPR. Microsoft 365 will be ready for May 25 this year when GDPR takes effect in Europe. We're committed to supporting and working closely with you to help accelerate your GDPR compliance.

By focusing on these four areas you can get the boost you need for your digital transformation. We're seeing companies from many industries evolving and innovating with Microsoft 365, and we're excited to see what you can do too. The modern workplace is here.

SCOM 1711 Technical Preview (1801?) New Features

$
0
0

 

What does the new SCOM bring?!

 

HTML5 Web Dashboards! here

The new SCOM version also gives visibility into

Management Packs Updates and Recommendations

Operations Manager Products (view SCOM topology)

Partner Solutions

 

For now I'll focus on the last two

 

Operations Manager Products

Click on Administration Tab

Expand Operations Manager Products

Topology features are pretty neat.

Whether you're new to SCOM, or have interesting Server naming conventions, or someone built the environment and changed jobs or left the company...

At least you can easily find out how the SCOM environment is setup

 

Partner Solutions

Cool!

Silect MP Author, Comtrade, Nutanix, Veeam, Infront, ClearPointe, Backbone

Replacement Azure Management Certificate Monitoring

$
0
0

We replaced one of the monitors for Azure Management Certificates, and I had a few requests to share it with the public.

We disabled this monitor:
- Management Pack: Microsoft.SystemCenter.WindowsAzure
- Monitor System Name: Microsoft.SystemCenter.WindowsAzure.Subscription.Certificate.Monitor
- Monitor Display Name: Windows Azure Subscription Certificate Expiration Monitor
- Monitor Description: Windows Azure Subscription Certificate Expiration Monitor
- Alert Title: Management Certificates will expire soon

We didn't like that monitor for several reasons. First, it doesn't say which certificates. It just lets you know that one or more are expired or expiring. And it doesn't give us a way to filter out known certificates that we don't want to monitor. For example, we had a large number of self-signed certs "CN=Windows Azure Tools" which did not need to be monitored.

So we created a very simple, replacement management pack, which I'm attaching to this post. It has one class, one relationship, one discovery, one monitor type (two alerts, one at 30 days and one at 7 days), and overrides the previously mentioned monitor.

The new class, Microsoft.SystemCenter.WindowsAzure.ManagementCertificate, has a separate instance for each certificate for each Azure subscription. So you can get separate alerts for separate certificates. Also, since the discovery is in PowerShell, in an unsealed MP, you can easily edit the logic to exclude certificates you don't want. Here's what we did:

If ( $myCert.Subject -ne 'CN=Windows Azure Tools' )
{

$myTimeSpan = New-TimeSpan $(Get-Date) $myCert.NotAfter
$DaysRemaining = $myTimeSpan.Days

$sInst = $discoveryData.CreateClassInstance("$MPElement[Name='Microsoft.SystemCenter.WindowsAzure.ManagementCertificate']$")

Obviously this is not an officially supported MP, so use at your own risk.

Download the Management Pack here: ManagementCertificates


Protected: Announcing: Transport Features and Performance Advancements in Fall Creators Update for Windows 10 and Windows Server 2016

$
0
0

This content is password protected. To view it please enter your password below:

Azureで新規作成したWindows Server 2008R2 SP1 イメージでWorkstationサービスが起動しない

$
0
0

こんにちは、Azureサポートチームの三國です。
今回はAzureで新規作成したWindows Server 2008R2 SP1 イメージでWorkstationサービスが起動しない際のトラブルシューティングについてご案内致します。

本情報の内容(添付文書、リンク先などを含む)は、作成日時点でのものであり、予告なく変更される場合があります。

発生条件


2018年頭以降にAzureのマーケットプレイスより作成されたWindows Server 2008R2 SP1イメージにおいて、Workstationサービスが起動しない事象が発生しています。以下に発生を確認する手順と原因を記載します。

[確認手順]

-Windows2008R2 SP1サーバにRDP接続する

-スタートメニューより"Service"を選択する

-"Workstation"サービスを右クリックし"Start"を選択する

-以下のメッセージが表示される場合は、今回のブログでご案内する事象と合致しています。

Windows could not start the Workstation service on Local Computer.
Error 1075: The dependency service does not exist or has been marked for deletion.

[原因]

MicrosoftがAzure Market Placeに発行しておりますWindows 2008R2 SP1イメージのバージョン"2.127.20171217"にて確認されるバグとなります。
なお、発行されているバージョンの一覧は以下のコマンドで確認が可能です。

PS C:Usersnamikun> Get-AzureRmVMImage -Location japaneast -PublisherName MicrosoftWindowsServer -Offer WindowsServer -sku "2008-R2-SP1"

Version        FilterExpression Skus        Offer         PublisherName          Location  Id
-------        ---------------- ----        -----         -------------          --------  --
2.127.20170406                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20170420                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20170510                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20170628                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20170712                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20170811                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20170918                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20171017                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20171115                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...
2.127.20171217                  2008-R2-SP1 WindowsServer MicrosoftWindowsServer japaneast /Subsc...

本事象はResource Manager Modelでのみで発生し、Classic Deploymentモデルでは発生しないことが確認できています。

対処策


正式な対処は、次期イメージの発行にてされます。次期イメージは2018年2月末頃に発行される予定です。

暫定的な回避としては、Azure PowerShellを用いてひとつ前の世代のイメージより仮想マシンを作成することが有効です。

Azure PowerShellより仮想マシンを作成する方法は以下のドキュメントをご参照ください。

<仮想マシンの作成 https://docs.microsoft.com/ja-jp/azure/virtual-machines/windows/tutorial-manage-vm >

上記手順の中では、仮想マシンのイメージを以下のコマンドで選択しています。

$vm = Set-AzureRmVMSourceImage `
    -VM $vm `
    -PublisherName MicrosoftWindowsServer `
    -Offer WindowsServer `
    -Skus 2016-Datacenter `
    -Version latest

そのコマンドを、以下のように変更することで、バージョン2.127.20171217の使用を回避できます。

$vm = Set-AzureRmVMSourceImage `
    -VM $vm `
    -PublisherName "MicrosoftWindowsServer" `
    -Offer "WindowsServer" `
    -Skus "2008-R2-SP1" `
    -Version "2.127.20171115"

その他のコマンドは先述したURLに従って実施することでWorkstationサービスの使用できるWin2008R2 SP1仮想マシンが作成できます。

---------

また、もう一つの対処方法として仮想マシンにて管理者権限を持つコマンドプロンプトより修正することもございます。手順は以下となります。

-仮想マシンにRDP接続する
-コマンドプロンプトを管理者権限で起動する
-以下のコマンドを打鍵する

sc config lanmanworkstation depend= bowser/mrxsmb20/nsi

以下の画像もご参照ください。

当コマンドを打鍵することで、Workstationサービスの起動が可能になります。

Creating custom authentication provider for OneDriveClient and GraphClient

$
0
0

This post is a contribution from Sohail Sayed, an engineer with the SharePoint Developer Support team

The following post demonstrates creating a custom authentication provider which can be used for authentication with the OneDriveClient or GraphClient classes.

 

Why create a custom authentication provider

The One Drive SDK sample at https://github.com/OneDrive/onedrive-sdk-csharp uses the MsaAuthenticationProvider which authenticates with OneDrive Personal only. There is an alternate DelegateAuthenticationProvider which works for OneDrive For Business but only for apps registered in https://apps.dev.microsoft.com and not in https://portal.azure.com .

Also you may scenarios where you may need to customize the authentication process. For example you may want to use a service account credential to authenticate and avoid prompting the user for credentials.

 

Creating the custom authentication provider

Creating a custom authentication provider involves the below steps

1.       Create a class implementing the Microsoft.Graph.IAuthenticationProvider

2.       Implement the method “Task AuthenticateRequestAsync(HttpRequestMessage request)” which will perform the authentication.

The AuthenticateRequestAsync method has a parameter of type HttpRequestMessage which represents the HttpRequest for the OneDriveClient or the GraphClient. The code for the authentication provider needs to add the AccessToken to the Authorization Header for this request.

 

Code Sample

Below implementation will prompt user for the first time to enter the user credentials for OneDrive. You need to reference the following NuGet Packages.

  • Microsoft.Graph.Core
  • Microsoft.IdentityModel.Clients.ActiveDirectory –  ADAL Library
using Microsoft.Graph;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.Net.Http;
using System.Security;
using Microsoft.IdentityModel.Clients.ActiveDirectory;

namespace Sample
{
    // simple implementation of ODBAuth
    public class ODBAuthentication : IAuthenticationProvider
    {

        internal const string GraphApiUrl = "https://graph.microsoft.com";
        internal const string Authority = "https://login.microsoftonline.com/";

        public string TenantId { get; private set; }

        public string ClientId { get; private set; }
        public string RedirectUrl { get; set; }
        public string AccessToken { get; private set; }

        private IntPtr handle;

        // this works with a windows application
        // We need to get the current window handle
        // Within a windows form you can get the current windows handle using the property this.Handle
        public ODBAuthentication(string tenantId, string clientId, string redirectUrl, IntPtr handle)
        {
            this.TenantId = tenantId;
            this.ClientId = clientId;
            this.RedirectUrl = redirectUrl;
            this.handle = handle;
        }

        AuthenticationResult authenticationResult = null;

        public async Task AuthenticateRequestAsync(HttpRequestMessage request)
        {
            AuthenticationContext authenticationContext = new AuthenticationContext(Authority + TenantId, false);

            try
            {
                // tracking if the user has signed atleast once within the app 
                // you can avoid this by using the option PromptBehavior.Auto but this will sign-in with the default user
                // use PromptBehavior.RefreshSession if you want to give user the option to choose the user login first time
                // Alternatively for console application or background applications use the overload of AcquireTokenAsync to signin with a service account using silent sign in
                if (authenticationResult == null)
                {
                    authenticationResult = await authenticationContext
                        .AcquireTokenAsync(GraphApiUrl, ClientId, new Uri(this.RedirectUrl), new PlatformParameters(PromptBehavior.RefreshSession, handle))
                        .ConfigureAwait(false);
                }
                else
                {
                    // Use PromptBehavior.Auto for subsequent requests
                    authenticationResult = await authenticationContext
                      .AcquireTokenAsync(GraphApiUrl, ClientId, new Uri(this.RedirectUrl), new PlatformParameters(PromptBehavior.Auto, handle))
                      .ConfigureAwait(false);
                }


                request.Headers.Add("Authorization", "Bearer " + authenticationResult.AccessToken);

            }
            catch (Exception ex)
            {
                throw ex;
            }
        }

    }
}

Below example uses the custom authentication provider with a OneDriveClient.  You need to create a native app in https://portal.azure.com with appropriate access for OneDrive and pass the TenantName, ClientID and RedirectUrl values from that app.

ODBAuthentication odbAuth = new ODBAuthentication(tenantName, clientId, redirectUrl, this.Handle);
var oneDriveClient = new OneDriveClient("https://graph.microsoft.com/v1.0/me/", odbAuth);

Using the custom authentication provider with GraphClient class is similar.

ODBAuthentication odbAuth = new ODBAuthentication(tenantName, clientId, redirectUrl, this.Handle);
var graphClient = new GraphServiceClient("https://graph.microsoft.com/v1.0",odbAuth);

文章から画像を生成するボットを開発

$
0
0

Posted by: ジョン ローチ (John Roach)

あなたが「黄色い胴体、黒い羽、短いくちばしの鳥を書いてください。」と言われたら、まず、鳥の輪郭を下描きし、黄色のペンで胴体を塗り、黒のペンで羽を塗って、最後にチェックをしてから、くちばしを短く描き、光の反射の表現を加えるでしょう。そして、鳥がとまっている木の枝も描くでしょう。

とうとう、ボットもこれと同じことができるようになりました。

Microsoft Researchで開発中の新しいAIテクノロジは、キャプションのような文章中の単語に注目して画像を生成します。arXiv.orgに投稿した研究論文のテストの結果では、このテクノロジは、従来のテキストから画像への変換手法の最先端技術と比較して、業界標準テストによる画像品質が約3倍に向上しています。

研究者が「ドローイングボット」と呼ぶこのテクノロジは、家畜の群れる農村の風景などの普通の景色から、宙に浮く2階建てバスといった馬鹿げたものまで、テキストからあらゆる画像を生成できます。各画像には、テキストの説明にはない詳細な背景などが含まれ、このAIには人工的な想像力を有していることを物語っています。

ワシントン州レドモンドのMicrosoft Research深層学習技術センターの主任研究員兼リサーチマネージャーであるチアドン・ホー (Xiaodong He) は次のように述べています。「Bingで鳥を検索すると鳥の画像が見つかりますが、このテクノロジを活用した場合、コンピューターによってゼロから鳥の画像が生成されます。生成された鳥は、現実には存在しないかもしれません。コンピューターが想像力によって生み出した鳥に過ぎないのです。」

ドローイングボットは、ホーと同僚たちが過去5年間にわたり研究してきた、画像と自然言語処理という両分野の交差によるものであり、それに関連する研究の集大成です。最初に、写真のキャプションを自動的に作成するテクノロジであるCaptionBotが作られ、次に、物体の位置や特性など画像に関する質問に回答するテクノロジが作られました(このテクノロジは視覚障がいを持つ人々にとって非常に有用です)。

これらの研究には物体を識別するための機械学習モデルの訓練、行動の解釈、そして、自然言語による対話が必要となります。

「今、私たちはテキストを使って画像を生成しようとしています。」と、ホーのグループのポストドクター研究者であり、論文“So, it is a cycle”(「なるほど、これが円ですね。」)の共著者であるチウユアン・ファン (Qiuyuan Huang) は言います。

画像の生成はキャプション付けよりも困難な作業です、とチームの一員であるペンツァン・チャン (Pengchuan Zhang)準研究員は付け加えます。これは、ドローイングボットがキャプションにない詳細部分の描写を作り上げなければならないためです。「機械学習アルゴリズムにより、画像中の欠けた部分を想像させなければならないのです。」

洞察力も備えた画像生成

マイクロソフトのドローイングボットの中核にあるのは GAN (Generative Adversarial Network:敵対的生成ネットワーク)と呼ばれるテクノロジです。このネットワークは 2つの機械学習モデルから構成され、生成ネットワークと呼ばれる一方のモデルが文章の記述から画像を生成し、もう一方の「ディスクリミネータ」と呼ばれるモデルが、生成された画像の正当性を文書の記述に基づいて判断します。生成ネットワークは間違った画像で識別ネットワークを欺こうとし、識別ネットワークは欺かれないように学習します。両者が協同することで、識別ネットワークは生成ネットワークを完璧な形に近づけていきます。

マイクロソフトのドローイングボットは画像とキャプションのペアから成るデータセットによって訓練しました。これにより、モデルは単語とその単語のビジュアルな表現との関係を学習することができます。GANは、たとえば、キャプションに鳥と書いてあった時には鳥を描くよう学習し、鳥の絵がどのようであるべきかを学習します。「これが、機械が学習できると私たちが考える根本的理由です」と、ホーは述べます。

GANは「青い鳥」や「常緑植物」といった単純な記述に基づいて画像を生成するのは得意ですが、「頭が緑色で、翼が黄色で、腹が赤色の鳥」といった複雑な記述になると画像の品質は低下します。これは、文章全体が生成ネットワークに対する単一の情報として扱われ、詳細な情報が失われてしまうからです。結果として得られるのは緑、黄色、赤が混ざったぼやけた鳥の画像であり、記述に合致するものではなくなります。

人間が絵を描く時にはテキストの記述を何回も参照し、今描いている部分に関する記述に対して注意を払います。この人間の特性を獲得するために、研究者たちは人間の洞察力を数学的に表現したAttnGAN(アテンショナル GAN)と呼ばれるネットワークを構築しました。このネットワークは、入力テキストを個別の単語に分解し、画像中の特定の領域と対応付けることで画像を生成します。

「洞察力は人間的概念です。私たちは数学を使ってコンピューターに洞察力を身につけさせます」とホーは説明します。

このモデルは人間が常識と考えるものを訓練データから学び、画像中の想像に任せられた部分に学習した概念を当てはめます。たとえば、訓練データ中の鳥の画像の多くは鳥が木の枝にとまっているため、AttnGANはテキストの記述と矛盾しない限り、鳥が木の上にとまっている画像を描きます。

「機械学習モデルは、鳥がどこにいるべきかという常識をデータから学習します」とチャンは言います。テストのために、チームは「赤い2階建てバスが湖に浮いている」という馬鹿げたキャプションをドローイングボットに入力しました。2つのデッキがあるボートと2階建てバスの両方に似た物体が山に囲まれた湖に浮かぶぼやけた画像が生成されました。これは、湖に浮かぶのはボートである一方でテキストではバスが指定されていることから、ボットが内部的に葛藤したことを示しています。

「記述を変更してマシンがどのように対応するかを見ることができます。そして対応の変化によりマシンが学習したことをテストできます。マシンはある程度の常識を学んでいますが、それでも矛盾する要求に従うことができ、時には馬鹿げた結果を生み出します」とホーは説明します。

実用的応用

テキストから画像への生成テクノロジには、画家やインテリアデザイナーのスケッチの補助や音声による写真編集ツールなどの実用的な応用が考えられます。ホーは、コンピューティング能力の強化により、台本に基づいたアニメーションの自動作成などを実用化できると構想しています。

現時点では、このテクノロジは完璧ではありません。画像をよく見ればほとんどの場合に欠陥があります。例えば、鳥のくちばしが黒ではなく青であったり、果物として売られているバナナが突然変異を起こしたようになっていたりします。これらの欠陥は、人間ではなくコンピューターが画像を作成していることの証拠でもあります。とは言え、AttnGANの画像の品質は、以前の最高水準のGANと比較して約3倍の向上を達成しています。これは、ホーによれば、人間の能力を高める汎用のインテリジェンスを実現する上での重要なマイルストーンです。

「AIと人間が同じ世界で共存するためには、互いにやり取りするための方法が必要です。そして、そのための最も重要な2つ方法が言語と画像です」とホーは説明します。

本研究には、マイクロソフトのチアドン・ホー、ペンツァン・チャン、チウユアン・ファンに加えて、インターンであるリーハイ大学のタオ・シュー(Tao Xu)、デューク大学のツェ・ガン (Zhe Gan) 、そして、ルトガー大学のハン・ツァ (Han Zhang)、リーハイ大学のシオライ・ファン (Xiaolei Huang)も協力しています。

 

関連資料(英語)

Read the research paper describing the AttnGAN
Learn more about Microsoft’s AI research efforts in Vision and Language Intelligence
Check out the CaptionBot and Seeing AI
Decades of computer vision research, one ‘Swiss Army knife’
John Roach writes about Microsoft research and innovation. Follow him on Twitter.

ーーー

本ページのすべての内容は、作成日時点でのものであり、予告なく変更される場合があります。正式な社内承認や各社との契約締結が必要な場合は、それまでは確定されるものではありません。また、様々な事由・背景により、一部または全部が変更、キャンセル、実現困難となる場合があります。予めご了承下さい。

Azure Discovery Days training and resources

$
0
0

On this blog we often talk about leveraging different Azure services alongside traditional on-premises capabilities, and if you haven't dipped your toes in the water yet it might be worth visiting one of the sessions that has been scheduled.

At the bottom of the post you will see links to some in-depth training opportunities as well.

Reserve your spot. Accelerate your learning.

Microsoft Azure is a growing collection of integrated cloud services – analytics, computing, database, storage, mobile, networking, the web – to help you move faster, work smarter and achieve more.

Its potential is unlimited if you know how to unlock it. Which is why we offer Azure Discovery Days; free sessions led by Microsoft Azure experts and built to help you extract extra value out of Azure.

PICK YOUR CITY. START YOUR JOURNEY.


ADELAIDE Azure Discovery Days 2018:

Adelaide

February 27th 2018 > Register Now

March 22nd 2018 > Register Now

 

 

 


BRISBANE Azure Discovery Days 2018:

Brisbane

February 13th 2018 > Register Now

March 13th 2018 > Register Now

 

 

 


CANBERRA Azure Discovery Days 2018:

Canberra

February 21st 2018 > Register Now

March 27th 2018 > Register Now

 

 

 


MELBOURNE Azure Discovery Days 2018:

Melbourne

January 30th 2018 > Register Now

February 12th 2018 > Register Now

February 14th 2018 > Register Now

March 13th 2018 > Register Now

March 14th 2018 > Register Now

 


PERTH Azure Discovery Days 2018:

Perth

February 27th 2018 > Register Now

March 27th 2018 > Register Now

March 28th 2018 > Register Now

 

 


SYDNEY Azure Discovery Days 2018:

Sydney

January 29th 2018 > Register Now

February 21st 2018 > Register Now

February 23rd > Register Now

March 22nd > Register Now

March 23rd > Register Now

 


Want to dive deeper?

After attending an Azure Discovery Day and gaining basic experience in implementing and monitoring Microsoft Azure solutions, you may be looking to dive deeper into the key capabilities of the Azure platform, how to choose the right data platform, computer platform and how to build and deploy applications using Azure.

Register for one of our more advanced 3-5 day Instructor led workshops.

*Course costs range from A$699 – A$899.

Course: Architecting Azure IaaS and Hybrid Solutions

Level: Advanced
Format: Instructor-led 3 day In-Person course
More information available here

Course: Architecture Big Data Analytics

Level: Advanced
Format: Instructor-led 3 day In-Person Course
More information available here

Course: Azure Fundamentals

Level: Beginner
Format: Instructor-led 2 day In-Person course
More information available here

Developing Microsoft Cloud Solutions with Azure.NET

Level: Advanced
Format: Instructor-led 3 day In-Person Course
More information available here

Implementing Microsoft Azure Infrastructure

Level: Advanced
Format: Instructor-led 5 day In-Person Course
More information available here

Introduction to Containers on Azure

Level: Intermediate
Format: Instructor-led 2 day In-Person Course
More information available here

Viewing all 36188 articles
Browse latest View live


<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>