研究开源源码之Myrmec

　　好久没写博客了，自己也弄不清是懒了还是忙了。毕竟白天需要工作，晚上有时候看看资料，有时候陪家人，有时候约朋友......更加累了，可能由于累了就懒得总结了。

　　今天有同事问我关于代码检查文件类型的问题。当然由于安全性不能直接使用文件后缀或者Mime检查，需要读取文件头，根据文件头来判断（作为判断的其中一种依据）。后来发现直接使用Myrmec是没有CSS和HTML文件检查的。不应该啊！无赖中只能上Github看看工具源码了。

　　Myrmec 是一个用于检测文件格式的库，Myrmec不同于其它库或者手写检测代码，Myrmec不依赖文件扩展名（在实际使用中，你的用户很可能使用虚假的扩展名欺骗你的应用程序），Myrmec会检测文件的二进制头，并在其元数据库中匹配来获得文件的格式。

　　例如Jpg图片的二进制头是 "FF D8 FF DB" 那么Myrmec会匹配到这个文件头，并获得两个结果--"jpg"和"jpeg"。

　　具体使用我就不详细说了，毕竟百度就有，而且比较简单（真的很简单，哈哈）。直接上源码地址https://github.com/rocketRobin/myrmec。

　　

　　代码其实也是比较简单的，其中FileTypes是已经标记为过期的类，里面定义了很多标记文件类型的16进制，用List<Record>存储。

　　

// <copyright file="FileTypes.cs" company="Rocket Robin">
// Copyright (c) Rocket Robin. All rights reserved.
// Licensed under the Apache v2 license. See LICENSE file in the project root for full license information.
// </copyright>

using System;
using System.Collections.Generic;

namespace Myrmec
{
    /// <summary>
    /// Common file types for populate a new sniffer instance.
    /// </summary>

    [Obsolete("please use populate the file types only you need.")]
    public class FileTypes
    {
        static FileTypes()
        {
            
            Unfrequent = new List<Record>
            {
                new Record("bin", "53 50 30 31"),
                new Record("bac", "42 41 43 4B 4D 49 4B 45 44 49 53 4B"),
                new Record("bz2", "42 5A 68"),
                new Record("tif tiff", "49 49 2A 00"),
                new Record("tif tiff", "4D 4D 00 2A"),
                new Record("cr2", "49 49 2A 00 10 00 00 00 43 52"),
                new Record("cin", "80 2A 5F D7"),
                new Record("exr", "76 2F 31 01"),
                new Record("dpx", "53 44 50 58"),
                new Record("dpx", "58 50 44 53"),
                new Record("bpg", "42 50 47 FB"),
                new Record("lz", "4C 5A 49 50"),
                new Record("ps", "25 21 50 53"),
                new Record("fits", "3D 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 54"),
                new Record("doc xls ppt msg", "D0 CF 11 E0 A1 B1 1A E1"),
                new Record("dex", "64 65 78 0A 30 33 35 00"),
                new Record("vmdk", "4B 44 4D"),
                new Record("crx", "43 72 32 34"),
                new Record("cwk", "05 07 00 00 42 4F 42 4F 05 07 00 00 00 00 00 00 00 00 00 00 00 01"),
                new Record("fh8", "41 47 44 33"),
                new Record("cwk", "06 07 E1 00 42 4F 42 4F 06 07 E1 00 00 00 00 00 00 00 00 00 00 01"),
                new Record("toast", "45 52 02 00 00 00"),
                new Record("toast", "8B 45 52 02 00 00 00"),
                new Record("xar", "78 61 72 21"),
                new Record("dat", "50 4D 4F 43 43 4D 4F 43"),
                new Record("nes", "4E 45 53 1A"),
                new Record("tox", "74 6F 78 33"),
                new Record("MLV", "4D 4C 56 49"),
                new Record("lz4", "04 22 4D 18"),
                new Record("cab", "4D 53 43 46"),
                new Record("flif", "46 4C 49 46"),
                new Record("stg", "4D 49 4C 20"),
                new Record("der", "30 82"),
                new Record("wasm", "00 61 73 6d"),
                new Record("lep", "cf 84 01"),
                new Record("rtf", "7B 5C 72 74 66 31"),
                new Record("m2p vob", "00 00 01 BA"),
                new Record("zlib", "78 01"),
                new Record("zlib", "78 9c"),
                new Record("zlib", "78 da"),
                new Record("lzfse", "62 76 78 32"),
                new Record("orc", "4F 52 43"),
                new Record("avro", "4F 62 6A 01"),
                new Record("rc", "53 45 51 36"),
                new Record("tbi", "00 00 00 00 14 00 00 00"),
                new Record("dat", "00 00 00 00 62 31 05 00 09 00 00 00 00 20 00 00 00 09 00 00 00 00 00 00", 8, "Bitcoin Core wallet.dat file"),
                new Record("jp2", "00 00 00 0C 6A 50 20 20 0D 0A", "Various JPEG-2000 image file formats"),
                new Record("ttf", "00 01 00 00 00"),
                new Record("mdf", "00 FF FF FF FF FF FF FF FF FF FF 00 00 02 00 01"),

                // Complex file type.
                new Record("pdb", "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00", 11),
                new Record("3gp 3g2", "66 74 79 70 33 67", 4),
                new Record("iso", "43 44 30 30 31", 32769),
                new Record("iso", "43 44 30 30 31", 34817),
                new Record("iso", "43 44 30 30 31", 36865),
            };
            Common = new List<Record>
            {
                new Record("asf wma wmv", "30 26 B2 75 8E 66 CF 11 A6 D9 00 AA 00 62 CE 6C"),
                new Record("ogg oga ogv", "4F 67 67 53"),
                new Record("psd", "38 42 50 53"),
                new Record("mp3", "FF FB"),
                new Record("mp3", "49 44 33"),
                new Record("bmp dib", "42 4D"),
                new Record("jpg,jpeg", "ff,d8,ff,db"),
                new Record("png", "89,50,4e,47,0d,0a,1a,0a"),
                new Record("zip,jar,odt,ods,odp,docx,xlsx,pptx,vsdx,apk,aar", "50,4b,03,04"),
                new Record("zip,jar,odt,ods,odp,docx,xlsx,pptx,vsdx,apk,aar", "50,4b,07,08"),
                new Record("zip,jar,odt,ods,odp,docx,xlsx,pptx,vsdx,apk,aar", "50,4b,05,06"),
                new Record("rar", "52,61,72,21,1a,07,00"),
                new Record("rar", "52,61,72,21,1a,07,01,00"),
                new Record("class", "CA FE BA BE"),
                new Record("pdf", "25 50 44 46"),
                new Record("rpm", "ed ab ee db"),
                new Record("flac", "66 4C 61 43"),
                new Record("mid midi", "4D 54 68 64"),
                new Record("ico", "00 00 01 00"),
                new Record("z,tar.z", "1F 9D"),
                new Record("z,tar.z", "1F A0"),
                new Record("gif", "47 49 46 38 37 61"),
                new Record("dmg", "78 01 73 0D 62 62 60"),
                new Record("gif", "47 49 46 38 39 61"),
                new Record("exe", "4D 5A"),
                new Record("tar", "75 73 74 61 72", 257),
                new Record("mkv mka mks mk3d webm", "1A 45 DF A3"),
                new Record("gz tar.gz", "1F 8B"),
                new Record("xz tar.xz", "FD 37 7A 58 5A 00 00"),
                new Record("7z", "37 7A BC AF 27 1C"),
                new Record("mpg mpeg", "00 00 01 BA"),
                new Record("mpg mpeg", "00 00 01 B3"),
                new Record("woff", "77 4F 46 46"),
                new Record("woff2", "77 4F 46 32"),
                new Record("XML", "3c 3f 78 6d 6c 20"),
                new Record("swf", "43 57 53"),
                new Record("swf", "46 57 53"),
                new Record("deb", "21 3C 61 72 63 68 3E"),

                // complext
                new Record("jpg,jpeg","FF D8 FF E0 ?? ?? 4A 46 49 46 00 01"),
                new Record("jpg,jpeg","FF D8 FF E1 ?? ?? 45 78 69 66 00 00"),
            };
        }

        /// <summary>
        /// Gets CommonFileTypes.
        /// Replace this with <see cref="Common"/>
        /// </summary>
        [Obsolete("please use populate the file types only you need.")]
        public static List<Record> CommonFileTypes { get => Common; }

        /// <summary>
        /// Gets Common It contains the format of the file we often see.
        /// </summary>
        [Obsolete("please use populate the file types only you need.")]
        public static List<Record> Common { get; set; }


        /// <summary>
        /// It contains unfrequent file formats.
        /// </summary>
        [Obsolete("please use populate the file types only you need.")]
        public static List<Record> Unfrequent { get; set; }
    }
}

　　Record类是用于存储文件类型和16进制对应。

// <copyright file="Record.cs" company="Rocket Robin">
// Copyright (c) Rocket Robin. All rights reserved.
// Licensed under the Apache v2 license. See LICENSE file in the project root for full license information.
// </copyright>

namespace Myrmec
{
    /// <summary>
    /// Present one record.
    /// </summary>
    public class Record
    {
        /// <summary>
        /// Initializes a new instance of the <see cref="Record"/> class.
        /// </summary>
        public Record()
        {
        }

        /// <summary>
        /// Initializes a new instance of the <see cref="Record"/> class.
        /// </summary>
        /// <param name="extentions">extentions string ,split with "," what if it has many.</param>
        /// <param name="hex">hex string, split with ",".</param>
        public Record(string extentions, string hex)
        {
            Extentions = extentions;
            Hex = hex;
        }

        /// <summary>
        /// Initializes a new instance of the <see cref="Record"/> class.
        /// </summary>
        /// <param name="extentions">Extentions format string.</param>
        /// <param name="hex">File hex head format string.</param>
        /// <param name="offset">Offset of this record.</param>
        public Record(string extentions, string hex, int offset)
        {
            Offset = offset;
            Extentions = extentions;
            Hex = hex;
        }

        /// <summary>
        /// Initializes a new instance of the <see cref="Record"/> class.
        /// </summary>
        /// <param name="extentions">extentions string ,split with "," what if it has many.</param>
        /// <param name="hex">hex string, split with ",".</param>
        /// <param name="description">description</param>
        public Record(string extentions, string hex, string description)
        {
            Extentions = extentions;
            Hex = hex;
            Description = description;
        }

        /// <summary>
        /// Initializes a new instance of the <see cref="Record"/> class.
        /// </summary>
        /// <param name="extentions">extentions string ,split with "," what if it has many.</param>
        /// <param name="hex">hex string, split with ",".</param>
        /// <param name="offset"></param>
        /// <param name="description">description</param>
        public Record(string extentions, string hex, int offset, string description)
        {
            Offset = offset;
            Extentions = extentions;
            Hex = hex;
            Description = description;
        }

        /// <summary>
        /// Gets or sets Description
        /// </summary>
        public string Description { get; set; }

        /// <summary>
        /// Gets or sets file extentions.
        /// </summary>
        public string Extentions { get; set; }

        /// <summary>
        /// Gets or sets Hex String.
        /// </summary>
        public string Hex { get; set; }

        /// <summary>
        /// Gets or sets offset
        /// </summary>
        public int Offset { get; set; }

        /// <summary>
        /// Gets a value indicating whether this record has offset or contain a variable byte or not.
        /// </summary>
        public bool IsComplexMetadata
        {
            get => (Offset > 0) || (Hex.Contains("?"));
        }
    }
}

　　Node类用于存储类型查询树，用于方便类型查询的。

// <copyright file="Node.cs" company="Rocket Robin">
// Copyright (c) Rocket Robin. All rights reserved.
// Licensed under the Apache v2 license. See LICENSE file in the project root for full license information.
// </copyright>

using System.Collections.Generic;

namespace Myrmec
{
    /// <summary>
    /// node
    /// </summary>
    public class Node
    {
        /// <summary>
        /// Initializes a new instance of the <see cref="Node"/> class.
        /// </summary>
        public Node()
        {
        }

        /// <summary>
        /// Gets or sets children.
        /// </summary>
        public SortedList<byte, Node> Children { get; set; }

        /// <summary>
        /// Gets or sets depth.
        /// </summary>
        public int Depth { get; set; }

        /// <summary>
        /// Gets or sets extentions.
        /// </summary>
        public List<string> Extentions { get; set; }

        /// <summary>
        /// Gets or sets parent node.
        /// </summary>
        public Node Parent { get; set; }
    }
}

　　Sniffer这个工具类是嗅探器用于提供根据文件16进制头判断类型等功能，很多核心算法都是在这个类中，后面我详细介绍这个算法

// <copyright file="Sniffer.cs" company="Rocket Robin">
// Copyright (c) Rocket Robin. All rights reserved.
// Licensed under the Apache v2 license. See LICENSE file in the project root for full license information.
// </copyright>

using System;
using System.Collections.Generic;
using System.Linq;

namespace Myrmec
{
    /// <summary>
    /// sniffer
    /// </summary>
    public class Sniffer
    {
        /// <summary>
        /// You can get the file extention name detail in this wikipedia page.
        /// </summary>
        public const string FileExtentionHelpUrl = "https://en.wikipedia.org/wiki/List_of_file_signatures";

        private Node _root;

        /// <summary>
        /// Initializes a new instance of the <see cref="Sniffer"/> class.
        /// </summary>
        public Sniffer()
        {
            _root = new Node()
            {
                Children = new SortedList<byte, Node>(128),
                Depth = -1,
            };
            ComplexMetadata = new List<Metadata>(10);
        }

        /// <summary>
        /// Gets or sets ComplexMetadatas.
        /// </summary>
        public List<Metadata> ComplexMetadata { get; set; }

        /// <summary>
        /// Add a record to matadata tree.
        /// </summary>
        /// <param name="data">file head.</param>
        /// <param name="extentions">file extention list.</param>
        public void Add(byte[] data, string[] extentions)
        {
            Add(data, _root, extentions, 0);
        }

        /// <summary>
        /// 
        /// </summary>
        /// <param name="record"></param>
        public void Add(Record record)
        {
            if (record.IsComplexMetadata)
            {
                ComplexMetadata.Add(record);
            }
            else
            {
                Add(record.Hex.GetByte(), record.Extentions.Split(',', ' '));
            }
        }

        /// <summary>
        /// Find extentions that match the file hex head.
        /// </summary>
        /// <param name="data">file hex head</param>
        /// <param name="matchAll">match all result or only the first.</param>
        /// <returns>matched result</returns>
        public List<string> Match(byte[] data, bool matchAll = false)
        {
            List<string> extentionStore = new List<string>(4);
            Match(data, 0, _root, extentionStore, matchAll);

            if (matchAll || !extentionStore.Any())
            {
                // Match data from complex metadata.
                extentionStore.AddRange(ComplexMetadata.Match(data, matchAll));
            }

            // Remove repeated extentions.
            if (matchAll && extentionStore.Any())
            {
                extentionStore = extentionStore.Distinct().ToList();
            }

            return extentionStore;
        }

        private void Add(byte[] data, Node parent, string[] extentions, int depth)
        {
            Node current = null;

            if (parent.Children == null)
            {
                parent.Children = new SortedList<byte, Node>(Convert.ToInt32(128 / Math.Pow(2, depth)));
            }

            // if not contains current byte index, create node and put it into children.
            if (!parent.Children.ContainsKey(data[depth]))
            {
                current = new Node
                {
                    Depth = depth,
                    Parent = parent
                };
                parent.Children.Add(data[depth], current);
            }
            else
            {
                if (!parent.Children.TryGetValue(data[depth], out current))
                {
                    throw new Exception("No possibility, something fucked up...");
                }

            }

            // last byte, put extentions into Extentions.
            if (depth == (data.Length - 1))
            {
                if (current.Extentions == null)
                {
                    current.Extentions = new List<string>(4);
                }

                current.Extentions.AddRange(extentions);
                return;
            }

            Add(data, current, extentions, depth + 1);
        }

        private void Match(byte[] data, int depth, Node node, List<string> extentionStore, bool matchAll)
        {
            // if depth out of data.Length's index then data end.
            if (data.Length == depth)
            {
                return;
            }

            node.Children.TryGetValue(data[depth], out Node current);

            // can't find matched node, match ended.
            if (current == null)
            {
                return;
            }

            // now extentions not null, this node is a final node and this is a result.
            if (current.Extentions != null)
            {
                extentionStore.AddRange(current.Extentions);

                // if only match first matched.
                if (!matchAll)
                {
                    return;
                }
            }

            // children is null, match ended.
            if (current.Children == null)
            {
                return;
            }

            // children not null, keep match.
            Match(data, depth + 1, current, extentionStore, matchAll);
        }
    }
}

　　好了，上面就是几个核心的类，现在讲解一下核心源码。类似这样的NUnit代码

[TestMethod]
        public void SnifferTest()
        {
            var sniffer = new Sniffer();
            sniffer.Populate(FileTypes.Common);
            sniffer.Populate(FileTypes.Unfrequent);
            var head = new byte[]
            {
                0xff,
                0xd8,
                0xff,
                0xdb
            };

            var result = sniffer.Match(head,true);

            Assert.IsTrue(result.Contains("jpg"));
            Assert.IsTrue(result.Contains("jpeg"));
        }

　　var sniffer = new Sniffer();这个是定义和初始化嗅探器，在这个过程中会使用Node类初始化树，这个树用于存储所有FileTypes下定义的节点。

public Sniffer()
        {
            _root = new Node()
            {
                Children = new SortedList<byte, Node>(128),
                Depth = -1,
            };
            ComplexMetadata = new List<Metadata>(10);
        }

　　sniffer.Populate(FileTypes.Common);这个作用是生成树。生成过程是一个递归算法

public static void Populate(this Sniffer sniffer, IList<Record> records)
        {
            foreach (var record in records)
            {
                sniffer.Add(record);
            }
        }

public void Add(Record record)
        {
            if (record.IsComplexMetadata)
            {
                ComplexMetadata.Add(record);
            }
            else
            {
                Add(record.Hex.GetByte(), record.Extentions.Split(',', ' '));
            }
        }

public void Add(byte[] data, string[] extentions)
        {
            Add(data, _root, extentions, 0);
        }

private void Add(byte[] data, Node parent, string[] extentions, int depth)
        {
            Node current = null;

            if (parent.Children == null)
            {
                parent.Children = new SortedList<byte, Node>(Convert.ToInt32(128 / Math.Pow(2, depth)));
            }

            // if not contains current byte index, create node and put it into children.
            if (!parent.Children.ContainsKey(data[depth]))
            {
                current = new Node
                {
                    Depth = depth,
                    Parent = parent
                };
                parent.Children.Add(data[depth], current);
            }
            else
            {
                if (!parent.Children.TryGetValue(data[depth], out current))
                {
                    throw new Exception("No possibility, something fucked up...");
                }

            }

            // last byte, put extentions into Extentions.
            if (depth == (data.Length - 1))
            {
                if (current.Extentions == null)
                {
                    current.Extentions = new List<string>(4);
                }

                current.Extentions.AddRange(extentions);
                return;
            }

            Add(data, current, extentions, depth + 1);
        }

　　生成树过程是首选把FileTypes.Common中预先定义Record列表的hex通过空格和逗号分隔，然后转成10进制，例如new Record("ogg oga ogv", "4F 67 67 53")。hex部分就是 "4F 67 67 53"，转换为10进制就是“79 103 103 83”，然后把这数字串生成树，生成图如下：

　　　如果再加上一个Record节点，如果有相同顺序部分，就使用相同的节点，例如再有一个路径是new Record("XXX", "4F 67 01")，此时树节点如下：

 　　明白了上面树结构，就很容易知道这树的作用了，就是为了提高查询的速度的。var result = sniffer.Match(head,true);就可以通过树的路径找到文件类型，即叶子节点对应的类型属性。

　　好了，这个工具其实原理很简单，就是使用的树结构提高了查询效率而已。

嗅探器