Configuring DDoS and BotNet Detector Based on QoE

The detector allows automatic monitoring of incoming traffic, identifying DDoS attacks, and quickly taking measures to protect the network. When a threat is detected, the system can fully isolate the malicious stream or clean the traffic while maintaining service availability for users.
The solution requires SSG version BASE, COMPLETE, or BNG with additional options and is deployed on an existing server with QoE.

On the QoE server.

Update QoE to the latest version, stopping the receivers beforehand. Before starting receivers, patch ClickHouse:

dnf --refresh install clickhouse-patched

Start the receivers.

On the GUI server.

Update GUI to the latest version. Connect GUI to VAS Cloud if not already connected. Enable the aniddos license option.

In the file /var/www/html/dpiui2/frontend/env.js set the option AppEnv.DDoSAttack_isVisible = 1;

On the QoE server.

Install the mitigator package fastm_qoe on all nodes:

dnf install fastm_qoe

Switch Python version:

dnf install -y python39 python39-devel -y
sudo update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.6 60
sudo update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.9 70
sudo update-alternatives --config python3

Select version 3.9:

python3 --version

On the QoE server.

On all nodes, or on selected ones:

1. Edit the file /var/fastm_qoe/etc/.env.
   It should contain the following:

   ANALYZER=avg-based-z-score
   ANALYZER_RULES_KEY=avg-based-z-score-any
   
   IDLE_MODE=1
   FORCE_MODE=0
   DB_DROP_TABLES=1
   
   FM_ATTACKS_METRICS_BY_SUBS_FILTER="and has_attack = 0"
   FM_ATTACKS_METRICS_BY_SUBS_LIMIT=1
   FM_ATTACKS_METRICS_BY_SUBS_COLLAPSE=1
   FM_ATTACKS_METRICS_BY_SUBS_DAY='day_'

2. Update schema:

   fastm-db-scheme

3. Enable metrics collection
   Add to file /var/qoestor/backend/.env the following:

   FM_FULLFLOW_HOOK_ENABLE=1
   GEO_IP_DIC_AUTOLOAD_ENABLED=1

   Execute the daily cron:

   sh /var/qoestor/backend/app_bash/cron_daily.sh

   Collect metrics for several hours, ideally 24 hours. Then edit /var/fastm_qoe/etc/.env again and change 2 parameters:

   IDLE_MODE=0
   DB_DROP_TABLES=0

   This activates the detector.

In the file /var/fastm_qoe/lib/rules/config.json, edit the section avg-based-z-score-any as follows:

"avg-based-z-score-any": {
        "octets": { "th": 100, "weight": 0.1 },
        "octets_dropped": { "th": 1000, "weight": 0.3 },
        "packets": { "th": 100, "weight": 0.3 },
        "packets_dropped": { "th": 1000, "weight": 0.3 },
        "flows": { "th": 100, "weight": 0.4 },
        "sessions": { "th": 100, "weight": 0.4 },
        "duration": { "th": 100, "weight": 0.01 },
        "host_ips": { "th": 100, "weight": 0.3 },
        "protos": { "th": 100, "weight": 0.3 },
        "bits_sec": { "th": 100, "weight": 0.05 },
        "bits_dropped_sec": { "th": 1000, "weight": 0.05 },
        "packets_sec": { "th": 100, "weight": 0.05 },
        "packets_dropped_sec": { "th": 1000, "weight": 0.05 }
    },

In the GUI web interface, configure storage of raw and aggregated metrics, as well as raw and aggregated attack logs.

In Admin → GUI Configuration → QoE Stor: set DB retention time values as follows:

• QOESTOR_FM_ATTACKS_MAIN_LOG_PARTITIONS_LIFE_TIME_HOUR = 720
• QOESTOR_FM_ATTACKS_AGG_LOG_PARTITIONS_LIFE_TIME_DAYS = 30
• QOESTOR_FM_METRICS_MAIN_LOG_PARTITIONS_LIFE_TIME_HOUR = 72
• QOESTOR_FM_METRICS_AGG_LOG_PARTITIONS_LIFE_TIME_DAYS = 7

Detected attacks can be examined in the DDoS attack sections in QoE Analytics.

1. Start with the "TOP Attacks" section for the last 24 hours.
   Sort attacks by number of sessions and note a few IPs with the highest session count.
   
2. Check the "TOP Attacks by Protocols" section
   Also sort by session count. Note these protocols.
3. Check the "TOP Attacking IP Addresses" section, note a few IPs with the highest session count
   
4. Check the Attack Log
   Filter by previously selected subscribers and protocol.
   Here you can get attack details and make additional conclusions to take actions.
   For example, the screenshot below clearly shows port scanning on the same address via UDP protocol. In this case, it is sufficient to place the attacker IP into a separate AS and set it to drop.