OSSIM Agent的主要职责是收集网络上存在的各种设备发送的所有数据,然后按照一种标准方式有序发给OSSIM Server,Agent收集到数据后在发送给Server之前要对这些数据进行归一化处理,本文主要就如何有序发送数据与如何完成归一化进行讨论。
OSSIM传感器在通过GET框架实现OSSIM代理和OSSIM服务器之间通信协议和数据格式的之间转换。下面我们先简要看一下ossim-agent脚本:

#!/usr/bin/python -OOt
import sys
sys.path.append('/usr/share/ossim-agent/')
sys.path.append('/usr/local/share/ossim-agent/')
from ossim_agent.Agent import Agent
agent = Agent()
agent.main()

这里需要GET作为OSSIM代理向OSSIM服务器输送数据。实现紧密整合所需的两个主要操作是“生成”(或)OSSIM兼容事件的“映射Mapping”)和此类数据向OSSIM的“传输”服务器。它负责此类操作的GET框架的两个组件是EventHandler和Sender Agent,如图1所示。
OSSIM传感器Agent传送机制初探
图1 将Get框架内容集成到OSSIM

Event Handler的主要任务是映射数据源插件采集的事件到SIEM实例警报的OSSIM标准化事件格式。为了执行这样的过程,原始消息经历由RAW LOG转换为现有归一化数据字段格式的一个转变;在上图中我们将这些机制表示为“归一化Normalization”和“OSSIM消息”。部分日志归一化代码:

from Logger import Logger
from time import mktime, strptime
logger = Logger.logger
class Event:
    EVENT_TYPE = 'event'
    EVENT_ATTRS = [
        "type",
        "date",
        "sensor",
        "interface",
        "plugin_id",
        "plugin_sid",
        "priority",
        "protocol",
        "src_ip",
        "src_port",
        "dst_ip",
        "dst_port",
        "username",
        "password",
        "filename",
        "userdata1",
        "userdata2",
        "userdata3",
        "userdata4",
        "userdata5",
        "userdata6",
        "userdata7",
        "userdata8",
        "userdata9",
        "occurrences",
        "log",
        "data",
        "snort_sid",    # snort specific
        "snort_cid",    # snort specific
        "fdate",
        "tzone"
    ]

    def __init__(self):
        self.event = {}
        self.event["event_type"] = self.EVENT_TYPE

    def __setitem__(self, key, value):

        if key in self.EVENT_ATTRS:
            self.event[key] = self.sanitize_value(value)
            if key == "date":
                # 以秒为单位
                self.event["fdate"]=self.event[key]
                try:
                    self.event["date"]=int(mktime(strptime(self.event[key],"%Y-%m-%d %H:%M:%S")))
                except:
                    logger.warning("There was an error parsing date (%s)" %\
                        (self.event[key]))
        elif key != 'event_type':
            logger.warning("Bad event attribute: %s" % (key))

    def __getitem__(self, key):
        return self.event.get(key, None)
    # 事件表示
    def __repr__(self):
        event = self.EVENT_TYPE
        for attr in self.EVENT_ATTRS:
            if self[attr]:
                event += ' %s="%s"' % (attr, self[attr])
        return event + "\n"
    # 返回内部哈希值
    def dict(self):
        return self.event

    def sanitize_value(self, string):
        return str(string).strip().replace("\"", "\\\"").replace("'", "")

class EventOS(Event):
    EVENT_TYPE = 'host-os-event'
    EVENT_ATTRS = [
        "host",
        "os",
        "sensor",
        "interface",
        "date",
        "plugin_id",
        "plugin_sid",
        "occurrences",
        "log",
        "fdate",
    ]

class EventMac(Event):
    EVENT_TYPE = 'host-mac-event'
    EVENT_ATTRS = [
        "host",
        "mac",
        "vendor",
        "sensor",
        "interface",
        "date",
        "plugin_id",
        "plugin_sid",
        "occurrences",
        "log",
        "fdate",
    ]

class EventService(Event):
    EVENT_TYPE = 'host-service-event'
    EVENT_ATTRS = [
        "host",
        "sensor",
        "interface",
        "port",
        "protocol",
        "service",
        "application",
        "date",
        "plugin_id",
        "plugin_sid",
        "occurrences",
        "log",
        "fdate",
    ]

class EventHids(Event):
    EVENT_TYPE = 'host-ids-event'
    EVENT_ATTRS = [
        "host",
        "hostname",
        "hids_event_type",
        "target",
        "what",
        "extra_data",
        "sensor",
        "date",
        "plugin_id",
        "plugin_sid",
        "username",
        "password",
        "filename",
        "userdata1",
        "userdata2",
        "userdata3",
        "userdata4",
        "userdata5",
        "userdata6",
        "userdata7",
        "userdata8",
        "userdata9",
        "occurrences",
        "log",
        "fdate",
    ]

class WatchRule(Event):

    EVENT_TYPE = 'event'
    EVENT_ATTRS = [
        "type",
    "date",
    "fdate",
    "sensor",
    "interface",
    "src_ip",
    "dst_ip",
    "protocol",
        "plugin_id",
        "plugin_sid",
        "condition",
        "value",
        "port_from",
        "src_port",
        "port_to",
        "dst_port",
        "interval",
        "from",
        "to",
        "absolute",
    "log",
        "userdata1",
        "userdata2",
        "userdata3",
        "userdata4",
        "userdata5",
        "userdata6",
        "userdata7",
        "userdata8",
        "userdata9",
        "filename",
        "username",
    ]
class Snort(Event):
    EVENT_TYPE = 'snort-event'
    EVENT_ATTRS = [
        "sensor",
        "interface",
        "gzipdata",
        "unziplen",
        "event_type",
        "plugin_id",
        "type",
        "occurrences"
    ]

日志编码代码:

import threading, time
from Logger import Logger
logger = Logger.logger
from Output import Output
import Config
import Event
from Threshold import EventConsolidation
from Stats import Stats
from ConnPro import ServerConnPro
class Detector(threading.Thread):
    def __init__(self, conf, plugin, conn):

        self._conf = conf
        self._plugin = plugin
        self.os_hash = {}
        self.conn = conn
        self.consolidation = EventConsolidation(self._conf)
        logger.info("Starting detector %s (%s).." % \
                    (self._plugin.get("config", "name"),
                     self._plugin.get("config", "plugin_id")))
        threading.Thread.__init__(self)
    def _event_os_cached(self, event):
        if isinstance(event, Event.EventOS):
            import string
            current_os = string.join(string.split(event["os"]), ' ')
            previous_os = self.os_hash.get(event["host"], '')
            if current_os == previous_os:
                return True
            else:
                # 失败并添加到缓存
                self.os_hash[event["host"]] = \
                    string.join(string.split(event["os"]), ' ')
        return False
    def _exclude_event(self, event):

        if self._plugin.has_option("config", "exclude_sids"):
            exclude_sids = self._plugin.get("config", "exclude_sids")
            if event["plugin_sid"] in Config.split_sids(exclude_sids):
                logger.debug("Excluding event with " +\
                    "plugin_id=%s and plugin_sid=%s" %\
                    (event["plugin_id"], event["plugin_sid"]))
                return True
        return False

    def _thresholding(self):
        self.consolidation.process()
    def _plugin_defaults(self, event):
        # 从配置文件中获取默认参数
        if self._conf.has_section("plugin-defaults"):

        # 1) 日期
            default_date_format = self._conf.get("plugin-defaults",
                                                 "date_format")
            if event["date"] is None and default_date_format and \
               'date' in event.EVENT_ATTRS:
                event["date"] = time.strftime(default_date_format, 
                                              time.localtime(time.time()))
        # 2) 传感器
            default_sensor = self._conf.get("plugin-defaults", "sensor")
            if event["sensor"] is None and default_sensor and \
               'sensor' in event.EVENT_ATTRS:
                event["sensor"] = default_sensor
        # 3) 网络接口
            default_iface = self._conf.get("plugin-defaults", "interface")
            if event["interface"] is None and default_iface and \
               'interface' in event.EVENT_ATTRS:
                event["interface"] = default_iface
        # 4) 源IP
            if event["src_ip"] is None and 'src_ip' in event.EVENT_ATTRS:
                event["src_ip"] = event["sensor"]
        # 5) 时区
            default_tzone = self._conf.get("plugin-defaults", "tzone")
            if event["tzone"] is None and 'tzone' in event.EVENT_ATTRS:
                event["tzone"] = default_tzone

        # 6) sensor,source ip and dest != localhost
            if event["sensor"] in ('127.0.0.1', '127.0.1.1'):
                event["sensor"] = default_sensor

            if event["dst_ip"] in ('127.0.0.1', '127.0.1.1'):
                event["dst_ip"] = default_sensor

            if event["src_ip"] in ('127.0.0.1', '127.0.1.1'):
                event["src_ip"] = default_sensor
        # 检测日志的类型
        if event["type"] is None and 'type' in event.EVENT_ATTRS:
            event["type"] = 'detector'
        return event
    def send_message(self, event):
        if self._event_os_cached(event):
            return

        if self._exclude_event(event):
            return
        #对于一些空属性使用默认值。
        event = self._plugin_defaults(event)

        # 合并之前检查
        if self.conn is not None:
            try:
                self.conn.send(str(event))
            except:
                id = self._plugin.get("config", "plugin_id")
                c = ServerConnPro(self._conf, id)
                self.conn = c.connect(0, 10)
                try:
                    self.conn.send(str(event))
                except:
                    return
            logger.info(str(event).rstrip())
        elif not self.consolidation.insert(event):
            Output.event(event)
        Stats.new_event(event)
    def stop(self):
        #self.consolidation.clear()
        pass
#在子类中重写
    def process(self):
        pass
    def run(self):
        self.process()
class ParserSocket(Detector):
    def process(self):
        self.process()
class ParserDatabase(Detector):
    def process(self):
        self.process()

… …

从上可以看出,传感器的归一化主要负责对每个LOG内数据字段进行重新编码,使其生成一个全新的可能用于发送到OSSIM服务器的完整事件。为达成这种目的GET框架中包含了一些特定的功能,以便将所有的功能转换需要BASE64转换的字段。“OSSIM消息”负责填充GET生成的原始事件中不存在的字段。所以上面讲的plugin_id、plugin_sid是用来表示日志消息来源类型和子类型,这也是生成SIEM事件的必填字段。为事件格式完整性考虑,有些时候在无法确认源或目标IP时,系统默认会采用0.0.0.0来填充该字段。

注意:这种必填字段我们可利用phpmyadmin工具查看OSSIM的MySQL数据库。
Sender Agent负责完成下面两个任务:
发送由GET收集并由事件格式化的事件发送到OSSIM服务器,这项任务由Event Hander创建的消息组成消息队列发送到消息中间件实现,时序图如图2所示。
OSSIM传感器Agent传送机制初探
图2 序列图: 从安全探测器的日志转换到OSSIM服务器事件

2)管理GET框架和OSSIM服务器之间的通信,通信端口为TCP 40001通过双向握手实现。归一化原始日志是规范化过程的一个重要环节,OSSIM在归一化处理日志的同时保留了原始日志,可用于日志归档,提供了一种从规范化事件中提取原始日志的手段。
经过归一化处理的EVENTS,存储到MySQL数据库中,如图3所示。接着就由关联引擎根据规则、优先级、可靠性等参数进行交叉关联分析,得出风险值并发出各种报警提示信息。
OSSIM传感器Agent传送机制初探
图3 OSSIM平台日志存储机制
接下来我们再看个实例,下面是一段Apache、CiscoASA以及SSH的原始日志,如图4、图5、图6所示。

  1. Apache插件中的正则表达式:
    [0001 - apache-access] 访问日志
    event_type=event
    regexp=((?P\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3})(:(?P\d{1,5}))? )?(?P\S+) (?P\S+) (?P\S+) \[(?P\d{2}\/\w{3}\/\d{4}:\d{2}:\d{2}:\d{2})\s+[+-]\d{4}\] \"(?P[^\"]*)\" (?P\d{3}) ((?P\d+)|-)( \"(?P[^\"]*)\" \"(?P[^\"]*)\")?$
    src_ip={resolv($src)}
    dst_ip={resolv($dst)}
    dst_port={$port}
    date={normalize_date($date)}
    plugin_sid={$code}
    username={$user}
    userdata1={$request}
    userdata2={$size}
    userdata3={$referer_uri}
    userdata4={$useragent}
    filename={$id}

[0002 - apache-error] 错误日志

event_type=event
regexp=\[(?P\w{3} \w{3} \d{2} \d{2}:\d{2}:\d{2} \d{4})\] \[(?P(emerg|alert|crit|error|warn|notice|info|debug))\] (\[client (?P\S+)\] )?(?P.*)
date={normalize_date($date)}
plugin_sid={translate($type)}
src_ip={resolv($src)}
userdata1={$data}

OSSIM传感器Agent传送机制初探
图4 Apache原始日志
OSSIM传感器Agent传送机制初探
图5 一条Cisco ASA 原始日志

OSSIM传感器Agent传送机制初探
图6 Cisco ASA 事件分类

通过过OSSIM归一化处理后的实际再通过Web前端展现给大家方便阅读的格式。归一化处理后的事件和原始日志的对比方法我们在《开源安全运维平台OSSIM疑难解析:入门篇》一书中还会讲解。而在图7所示的例子当中,仅使用了Userdata1和Userdata2,并没有用到Userdata3~Userdata9这些是扩展位,主要是为了预留给其他设备或服务使用,这里目标地址会标记成IP地址的形式,例如:Host192.168.11.160。实际上归一化处理这种操作发生在系统采集和存储事件之后,关联和数据分析之前,在SIEM工具中把采集过程中把数据转换成易读懂的格式,采用格式化的数据,能更容易理解。
OSSIM传感器Agent传送机制初探