Ookla: rework detection

The logic of the LRU cache has been changed: once we know an ip has
connected to an Ookla server, all the following (unknown) flows (for
a short time interval) from the same ip to the port 8080 are treated
as Ookla ones.

Most of the changes in this commit are about introducing the concept of
"aggressive detection". In some cases, to properly detect a
protocol we might use some statistical/behavior logic that, from one
side, let us to identify the protocol more often but, from the other
side, might lead to some false positives.
To allow the user/application to easily detect when such logic has been
triggered, the new confidence value `NDPI_CONFIDENCE_DPI_AGGRESSIVE` has been
added.
It is always possible to disable/configure this kind of logic via the
API.

Detection of Ookla flows using plain TLS over port 8080 is the first
example of aggressive detection in nDPI.

Tested with:
* Android 9.0 with app 4.8.3
* Ubuntu 20.04 with Firefox 110
* Win 10 with app 1.15 and 1.16
* Win 10 with Chrome 108, Edge 108 and Firefox 106

example/ndpiReader.c

@@ -96,6 +96,7 @@ u_int8_t verbose = 0, enable_flow_stats = 0;
 int nDPI_LogLevel = 0;
 char *_debug_protocols = NULL;
 char *_disabled_protocols = NULL;
+int aggressiveness[NDPI_MAX_SUPPORTED_PROTOCOLS];
 static u_int8_t stats_flag = 0;
 ndpi_init_prefs init_prefs = ndpi_no_prefs;
 u_int8_t human_readeable_string_len = 5;
@@ -515,6 +516,7 @@ static void help(u_int long_help) {
          "  -z                        | Enable JA3+\n"
          "  -A                        | Dump internal statistics (LRU caches / Patricia trees / Ahocarasick automas / ...\n"
          "  -M                        | Memory allocation stats on data-path (only by the library). It works only on single-thread configuration\n"
+         "  -Z proto:value            | Set this value of aggressiveness for this protocol (0 to disable it). This flag can be used multiple times\n"
          ,
          human_readeable_string_len,
          min_pattern_len, max_pattern_len, max_num_packets_per_flow, max_packet_payload_dissection,
@@ -797,7 +799,7 @@ void printCSVHeader() {
  */
 static void parseOptions(int argc, char **argv) {
   int option_idx = 0;
-  int opt;
+  int opt, i;
 #ifndef USE_DPDK
   char *__pcap_file = NULL;
   int thread_id, do_capture = 0;
@@ -818,7 +820,10 @@ static void parseOptions(int argc, char **argv) {
   }
 #endif
 
-  while((opt = getopt_long(argc, argv, "a:Ab:B:e:Ec:C:dDf:g:i:Ij:k:K:S:hHp:pP:l:r:s:tu:v:V:n:rp:x:w:zq0123:456:7:89:m:MT:U:",
+  for(i = 0; i < NDPI_MAX_SUPPORTED_PROTOCOLS; i++)
+    aggressiveness[i] = -1; /* Use the default value */
+
+  while((opt = getopt_long(argc, argv, "a:Ab:B:e:Ec:C:dDf:g:i:Ij:k:K:S:hHp:pP:l:r:s:tu:v:V:n:rp:x:w:zZ:q0123:456:7:89:m:MT:U:",
                            longopts, &option_idx)) != EOF) {
 #ifdef DEBUG_TRACE
     if(trace) fprintf(trace, " #### Handling option -%c [%s] #### \n", opt, optarg ? optarg : "");
@@ -950,6 +955,35 @@ static void parseOptions(int argc, char **argv) {
       _disabled_protocols = ndpi_strdup(optarg);
       break;
 
+    case 'Z': /* proto_name:aggr_value */
+      {
+        struct ndpi_detection_module_struct *module_tmp;
+        NDPI_PROTOCOL_BITMASK all;
+        char *saveptr, *tmp_str, *proto_str, *aggr_str;
+
+        /* Use a temporary module with all protocols enabled */
+        module_tmp = ndpi_init_detection_module(0);
+        if(!module_tmp)
+          break;
+        NDPI_BITMASK_SET_ALL(all);
+        ndpi_set_protocol_detection_bitmask2(module_tmp, &all);
+        ndpi_finalize_initialization(module_tmp);
+
+        tmp_str = ndpi_strdup(optarg);
+        if(tmp_str) {
+          proto_str = strtok_r(tmp_str, ":", &saveptr);
+          if(proto_str) {
+            aggr_str = strtok_r(NULL, ":", &saveptr);
+            if(aggr_str) {
+              aggressiveness[ndpi_get_protocol_id(module_tmp, proto_str)] = atoi(aggr_str);
+            }
+          }
+        }
+        ndpi_free(tmp_str);
+        ndpi_exit_detection_module(module_tmp);
+        break;
+      }
+
     case 'h':
       help(0);
       break;
@@ -2413,6 +2447,7 @@ static void debug_printf(u_int32_t protocol, void *id_struct,
 static void setupDetection(u_int16_t thread_id, pcap_t * pcap_handle) {
   NDPI_PROTOCOL_BITMASK enabled_bitmask;
   struct ndpi_workflow_prefs prefs;
+  int i;
 
   memset(&prefs, 0, sizeof(prefs));
   prefs.decode_tunnels = decode_tunnels;
@@ -2472,6 +2507,12 @@ static void setupDetection(u_int16_t thread_id, pcap_t * pcap_handle) {
 			  NDPI_LRUCACHE_BITTORRENT, 32768);
   /* Enable/disable LRU caches TTL here */
 
+  /* Set aggressiviness here */
+  for(i = 0; i < NDPI_MAX_SUPPORTED_PROTOCOLS; i++) {
+    if(aggressiveness[i] != -1)
+      ndpi_set_protocol_aggressiveness(ndpi_thread_info[thread_id].workflow->ndpi_struct, i, aggressiveness[i]);
+  }
+
   ndpi_finalize_initialization(ndpi_thread_info[thread_id].workflow->ndpi_struct);
 
   if(enable_doh_dot_detection)

fuzz/fuzz_config.cpp

@@ -108,6 +108,11 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   ndpi_set_opportunistic_tls(ndpi_info_mod, random_proto, random_value);
   ndpi_get_opportunistic_tls(ndpi_info_mod, random_proto);
 
+  for(i = 0; i < NDPI_MAX_SUPPORTED_PROTOCOLS; i++) {
+    ndpi_set_protocol_aggressiveness(ndpi_info_mod, i, random_value);
+    ndpi_get_protocol_aggressiveness(ndpi_info_mod, i);
+  }
+
   ndpi_finalize_initialization(ndpi_info_mod);
 
   /* Random protocol configuration */

src/include/ndpi_api.h

@@ -1031,6 +1031,11 @@ extern "C" {
   int ndpi_get_opportunistic_tls(struct ndpi_detection_module_struct *ndpi_struct,
 				 u_int16_t proto);
 
+  int ndpi_set_protocol_aggressiveness(struct ndpi_detection_module_struct *ndpi_struct,
+                                       u_int16_t proto, u_int32_t value);
+  u_int32_t ndpi_get_protocol_aggressiveness(struct ndpi_detection_module_struct *ndpi_struct,
+                                             u_int16_t proto);
+
   /**
    * Find a protocol id associated with a string automata
    *

src/include/ndpi_typedefs.h

@@ -659,6 +659,14 @@ struct ndpi_lru_cache {
   struct ndpi_lru_cache_entry *entries;
 };
 
+
+/* Aggressiveness values */
+
+#define NDPI_AGGRESSIVENESS_DISABLED			0x00 /* For all protocols */
+
+/* Ookla */
+#define NDPI_AGGRESSIVENESS_OOKLA_TLS			0x01 /* Enable detection over TLS (using ookla cache) */
+
 /* ************************************************** */
 
 struct ndpi_flow_tcp_struct {
@@ -920,7 +928,8 @@ typedef enum {
   NDPI_CONFIDENCE_DPI_CACHE,                /* Classification results based on some LRU cache (i.e. correlation among sessions) */
   NDPI_CONFIDENCE_DPI,                      /* Deep packet inspection */
   NDPI_CONFIDENCE_MATCH_BY_IP,              /* Classification obtained looking only at the IP addresses */
-
+  NDPI_CONFIDENCE_DPI_AGGRESSIVE,           /* Aggressive DPI: it might be a false positive */
+
   /*
     IMPORTANT
 
@@ -1229,6 +1238,8 @@ struct ndpi_detection_module_struct {
   int opportunistic_tls_pop_enabled;
   int opportunistic_tls_ftp_enabled;
 
+  u_int32_t aggressiveness_ookla;
+
   u_int16_t ndpi_to_user_proto_id[NDPI_MAX_NUM_CUSTOM_PROTOCOLS]; /* custom protocolId mapping */
   ndpi_proto_defaults_t proto_defaults[NDPI_MAX_SUPPORTED_PROTOCOLS+NDPI_MAX_NUM_CUSTOM_PROTOCOLS];
 
@@ -1563,6 +1574,10 @@ struct ndpi_flow_struct {
   /* NDPI_PROTOCOL_Z3950 */
   u_int8_t z3950_stage : 2; // 0-3
 
+  /* NDPI_PROTOCOL_OOKLA */
+  u_int8_t ookla_stage : 1;
+
+
   /* NDPI_PROTOCOL_OPENVPN */
   u_int8_t ovpn_session_id[8];
   u_int8_t ovpn_counter;

src/lib/ndpi_main.c

@@ -198,6 +198,10 @@ extern void ndpi_unset_risk(struct ndpi_detection_module_struct *ndpi_str,
 			    struct ndpi_flow_struct *flow, ndpi_risk_enum r);
 extern u_int32_t make_mining_key(struct ndpi_flow_struct *flow);
 extern int stun_search_into_zoom_cache(struct ndpi_detection_module_struct *ndpi_struct, struct ndpi_flow_struct *flow);
+extern void ookla_add_to_cache(struct ndpi_detection_module_struct *ndpi_struct,
+                               struct ndpi_flow_struct *flow);
+extern int ookla_search_into_cache(struct ndpi_detection_module_struct *ndpi_struct,
+                                   struct ndpi_flow_struct *flow);
 
 /* Forward */
 static int addDefaultPort(struct ndpi_detection_module_struct *ndpi_str,
@@ -2932,7 +2936,7 @@ struct ndpi_detection_module_struct *ndpi_init_detection_module(ndpi_init_prefs
   ndpi_str->msteams_cache_num_entries = 1024;
   ndpi_str->stun_zoom_cache_num_entries = 1024;
 
-  ndpi_str->ookla_cache_ttl = 0;
+  ndpi_str->ookla_cache_ttl = 120; /* sec */
   ndpi_str->bittorrent_cache_ttl = 0;
   ndpi_str->zoom_cache_ttl = 0;
   ndpi_str->stun_cache_ttl = 0;
@@ -2946,6 +2950,8 @@ struct ndpi_detection_module_struct *ndpi_init_detection_module(ndpi_init_prefs
   ndpi_str->opportunistic_tls_pop_enabled = 1;
   ndpi_str->opportunistic_tls_ftp_enabled = 1;
 
+  ndpi_str->aggressiveness_ookla = NDPI_AGGRESSIVENESS_OOKLA_TLS;
+
   for(i = 0; i < NUM_CUSTOM_CATEGORIES; i++)
     ndpi_snprintf(ndpi_str->custom_category_labels[i], CUSTOM_CATEGORY_LABEL_LEN, "User custom category %u",
 	     (unsigned int) (i + 1));
@@ -6254,6 +6260,13 @@ ndpi_protocol ndpi_detection_giveup(struct ndpi_detection_module_struct *ndpi_st
     ret.app_protocol = flow->detected_protocol_stack[0];
   }
 
+  /* Does it looks like Ookla? */
+  if(ret.app_protocol == NDPI_PROTOCOL_UNKNOWN &&
+     ntohs(flow->s_port) == 8080 && ookla_search_into_cache(ndpi_str, flow)) {
+    ndpi_set_detected_protocol(ndpi_str, flow, NDPI_PROTOCOL_OOKLA, NDPI_PROTOCOL_UNKNOWN, NDPI_CONFIDENCE_DPI_PARTIAL_CACHE);
+    ret.app_protocol = flow->detected_protocol_stack[0];
+  }
+
   /* Classification by-port is the last resort */
   if(enable_guess && ret.app_protocol == NDPI_PROTOCOL_UNKNOWN) {
 
@@ -8052,6 +8065,9 @@ const char *ndpi_confidence_get_name(ndpi_confidence_t confidence)
   case NDPI_CONFIDENCE_MATCH_BY_IP:
     return "Match by IP";
 
+  case NDPI_CONFIDENCE_DPI_AGGRESSIVE:
+    return "DPI (aggressive)";
+
   default:
     return NULL;
   }
@@ -8572,6 +8588,11 @@ int ndpi_match_hostname_protocol(struct ndpi_detection_module_struct *ndpi_struc
     ndpi_set_detected_protocol(ndpi_struct, flow, subproto, master_protocol, NDPI_CONFIDENCE_DPI);
     if(!category_depends_on_master(master_protocol))
       ndpi_int_change_category(ndpi_struct, flow, ret_match.protocol_category);
+
+    if(subproto == NDPI_PROTOCOL_OOKLA) {
+	ookla_add_to_cache(ndpi_struct, flow);
+    }
+
     return(1);
   } else
     return(0);
@@ -9643,3 +9664,36 @@ int ndpi_get_opportunistic_tls(struct ndpi_detection_module_struct *ndpi_struct,
     return -1;
   }
 }
+
+/* ******************************************************************** */
+
+int ndpi_set_protocol_aggressiveness(struct ndpi_detection_module_struct *ndpi_struct,
+                                     u_int16_t proto, u_int32_t value)
+{
+  if(!ndpi_struct)
+    return -1;
+
+  switch(proto) {
+  case NDPI_PROTOCOL_OOKLA:
+    ndpi_struct->aggressiveness_ookla = value;
+    return 0;
+  default:
+    return -1;
+  }
+}
+
+/* ******************************************************************** */
+
+u_int32_t ndpi_get_protocol_aggressiveness(struct ndpi_detection_module_struct *ndpi_struct,
+                                           u_int16_t proto)
+{
+  if(!ndpi_struct)
+    return -1;
+
+  switch(proto) {
+  case NDPI_PROTOCOL_OOKLA:
+    return ndpi_struct->aggressiveness_ookla;
+  default:
+    return -1;
+  }
+}

src/lib/protocols/http.c

@@ -46,6 +46,9 @@ static const char* binary_file_ext[] = {
 					NULL
 };
 
+extern void ookla_add_to_cache(struct ndpi_detection_module_struct *ndpi_struct,
+                               struct ndpi_flow_struct *flow);
+
 static void ndpi_search_http_tcp(struct ndpi_detection_module_struct *ndpi_struct,
 				 struct ndpi_flow_struct *flow);
 
@@ -436,6 +439,7 @@ static void ndpi_http_parse_subprotocol(struct ndpi_detection_module_struct *ndp
           || (strstr(flow->http.url, ":8080/upload?n=0.") != NULL))) {
 	/* This looks like Ookla speedtest */
 	ndpi_set_detected_protocol(ndpi_struct, flow, NDPI_PROTOCOL_OOKLA, master_protocol, NDPI_CONFIDENCE_DPI);
+	ookla_add_to_cache(ndpi_struct, flow);
       }
     }
 
@@ -1217,30 +1221,6 @@ static void ndpi_check_http_tcp(struct ndpi_detection_module_struct *ndpi_struct
         return;
       }
 
-      if((packet->payload_packet_len == 3) && memcmp(packet->payload, "HI\n", 3) == 0) {
-	/* This looks like Ookla: we don't give up with HTTP yet */
-        flow->l4.tcp.http_stage = 1;
-	return;
-      }
-
-      if((packet->payload_packet_len == 40) && (flow->l4.tcp.http_stage == 0)) {
-        /*
-	  -> QR O06L0072-6L91-4O43-857J-K8OO172L6L51
-	  <- QNUUX 2.5 2017-08-15.1314.4jn12m5
-	  -> MXFWUXJM 31625365
-	*/
-
-        if((packet->payload[2] == ' ')
-	   && (packet->payload[11] == '-')
-	   && (packet->payload[16] == '-')
-	   && (packet->payload[21] == '-')
-	   && (packet->payload[26] == '-')
-	   && (packet->payload[39] == 0x0A)
-	   )
-	  flow->l4.tcp.http_stage = 1;
-	return;
-      }
-
       if((packet->payload_packet_len == 23) && (memcmp(packet->payload, "<policy-file-request/>", 23) == 0)) {
         /*
           <policy-file-request/>
@@ -1251,25 +1231,7 @@ static void ndpi_check_http_tcp(struct ndpi_detection_module_struct *ndpi_struct
         */
       ookla_found:
         ndpi_int_http_add_connection(ndpi_struct, flow, NDPI_PROTOCOL_OOKLA, NDPI_PROTOCOL_CATEGORY_WEB);
-
-	if(ndpi_struct->ookla_cache != NULL) {
-	  if(packet->iph != NULL) {
-	    if(packet->tcp->source == htons(8080))
-	      ndpi_lru_add_to_cache(ndpi_struct->ookla_cache, packet->iph->saddr, 1 /* dummy */, ndpi_get_current_time(flow));
-	    else
-	      ndpi_lru_add_to_cache(ndpi_struct->ookla_cache, packet->iph->daddr, 1 /* dummy */, ndpi_get_current_time(flow));
-	  } else if(packet->iphv6 != NULL) {
-	    u_int32_t h;
-
-	    if(packet->tcp->source == htons(8080))
-	      h = ndpi_quick_hash((unsigned char *)&packet->iphv6->ip6_src, sizeof(packet->iphv6->ip6_src));
-	    else
-	      h = ndpi_quick_hash((unsigned char *)&packet->iphv6->ip6_dst, sizeof(packet->iphv6->ip6_dst));
-
-	    ndpi_lru_add_to_cache(ndpi_struct->ookla_cache, h, 1 /* dummy */, ndpi_get_current_time(flow));
-	  }
-	}
-
+        ookla_add_to_cache(ndpi_struct, flow);
         return;
       }
 
@@ -1389,18 +1351,6 @@ static void ndpi_check_http_tcp(struct ndpi_detection_module_struct *ndpi_struct
   } else if((flow->l4.tcp.http_stage == 1) || (flow->l4.tcp.http_stage == 2)) {
     NDPI_LOG_DBG2(ndpi_struct, "HTTP stage %u: \n", flow->l4.tcp.http_stage);
 
-    if((packet->payload_packet_len == 34) && (flow->l4.tcp.http_stage == 1)) {
-      if((packet->payload[5] == ' ') && (packet->payload[9] == ' ')) {
-	goto ookla_found;
-      }
-    }
-
-    if((packet->payload_packet_len > 6) && memcmp(packet->payload, "HELLO ", 6) == 0) {
-      /* This looks like Ookla */
-      goto ookla_found;
-    } else
-      NDPI_ADD_PROTOCOL_TO_BITMASK(flow->excluded_protocol_bitmask, NDPI_PROTOCOL_OOKLA);
-
     /**
        At first check, if this is for sure a response packet
        (in another direction. If not, if HTTP is detected do nothing now and return,