1 package edu.uci.iotproject.analysis;
3 import edu.uci.iotproject.Conversation;
4 import edu.uci.iotproject.DnsMap;
5 import edu.uci.iotproject.util.PcapPacketUtils;
6 import org.pcap4j.core.PcapPacket;
7 import org.pcap4j.packet.IpV4Packet;
8 import org.pcap4j.packet.TcpPacket;
11 import java.util.stream.Collectors;
12 import java.util.stream.Stream;
14 import static edu.uci.iotproject.util.PcapPacketUtils.*;
17 * Utility functions for analyzing and structuring (sets of) {@link Conversation}s.
19 * @author Janus Varmarken {@literal <jvarmark@uci.edu>}
20 * @author Rahmadi Trimananda {@literal <rtrimana@uci.edu>}
22 public class TcpConversationUtils {
26 * Given a {@link Conversation}, extract its set of "packet pairs", i.e., pairs of request-reply packets.
27 * <em>The extracted pairs are formed from the full set of payload-carrying TCP packets.</em>
30 * <b>Note:</b> in the current implementation, if one endpoint sends multiple packets back-to-back with no
31 * interleaved reply packets from the other endpoint, such packets are converted to one-item pairs (i.e., instances
32 * of {@link PcapPacketPair} where {@link PcapPacketPair#getSecond()} is {@code null}).
34 * @param conv The {@code Conversation} for which packet pairs are to be extracted.
35 * @return The packet pairs extracted from {@code conv}.
37 public static List<PcapPacketPair> extractPacketPairs(Conversation conv) {
38 return extractPacketPairs(conv.getPackets());
44 * Given a {@link Conversation}, extract its set of "packet pairs", i.e., pairs of request-reply packets.
45 * <em>The extracted pairs are formed from the full set of TLS Application Data packets.</em>
48 * <b>Note:</b> in the current implementation, if one endpoint sends multiple packets back-to-back with no
49 * interleaved reply packets from the other endpoint, such packets are converted to one-item pairs (i.e., instances
50 * of {@link PcapPacketPair} where {@link PcapPacketPair#getSecond()} is {@code null}).
52 * @param conv The {@code Conversation} for which packet pairs are to be extracted.
53 * @return The packet pairs extracted from {@code conv}.
55 public static List<PcapPacketPair> extractTlsAppDataPacketPairs(Conversation conv) {
57 throw new IllegalArgumentException(String.format("Provided %s argument is not a TLS session"));
59 return extractPacketPairs(conv.getTlsApplicationDataPackets());
62 // Helper method for implementing the public API of similarly named methods.
63 private static List<PcapPacketPair> extractPacketPairs(List<PcapPacket> packets) {
64 List<PcapPacketPair> pairs = new ArrayList<>();
66 while (i < packets.size()) {
67 PcapPacket p1 = packets.get(i);
68 String p1SrcIp = p1.get(IpV4Packet.class).getHeader().getSrcAddr().getHostAddress();
69 int p1SrcPort = p1.get(TcpPacket.class).getHeader().getSrcPort().valueAsInt();
70 if (i+1 < packets.size()) {
71 PcapPacket p2 = packets.get(i+1);
72 if (PcapPacketUtils.isSource(p2, p1SrcIp, p1SrcPort)) {
73 // Two packets in a row going in the same direction -> create one item pair for p1
74 pairs.add(new PcapPacketPair(p1, null));
75 // Advance one packet as the following two packets may form a valid two-item pair.
78 // The two packets form a response-reply pair, create two-item pair.
79 pairs.add(new PcapPacketPair(p1, p2));
80 // Advance two packets as we have already processed the packet at index i+1 in order to create the pair.
84 // Last packet of conversation => one item pair
85 pairs.add(new PcapPacketPair(p1, null));
86 // Advance i to ensure termination.
91 // TODO: what if there is long time between response and reply packet? Should we add a threshold and exclude those cases?
95 * Given a collection of TCP conversations and associated DNS mappings, groups the conversations by hostname.
96 * @param tcpConversations The collection of TCP conversations.
97 * @param ipHostnameMappings The associated DNS mappings.
98 * @return A map where each key is a hostname and its associated value is a list of conversations where one of the
99 * two communicating hosts is that hostname (i.e. its IP maps to the hostname).
101 public static Map<String, List<Conversation>> groupConversationsByHostname(Collection<Conversation> tcpConversations, DnsMap ipHostnameMappings) {
102 HashMap<String, List<Conversation>> result = new HashMap<>();
103 for (Conversation c : tcpConversations) {
104 if (c.getPackets().size() == 0) {
105 String warningStr = String.format("Detected a %s [%s] with no payload packets.",
106 c.getClass().getSimpleName(), c.toString());
107 System.err.println(warningStr);
110 IpV4Packet firstPacketIp = c.getPackets().get(0).get(IpV4Packet.class);
111 String ipSrc = firstPacketIp.getHeader().getSrcAddr().getHostAddress();
112 String ipDst = firstPacketIp.getHeader().getDstAddr().getHostAddress();
113 // Check if src or dst IP is associated with one or more hostnames.
114 Set<String> hostnames = ipHostnameMappings.getHostnamesForIp(ipSrc);
115 if (hostnames == null) {
116 // No luck with src ip (possibly because it's a client->srv packet), try dst ip.
117 hostnames = ipHostnameMappings.getHostnamesForIp(ipDst);
119 if (hostnames != null) {
120 // Put a reference to the conversation for each of the hostnames that the conversation's IP maps to.
121 for (String hostname : hostnames) {
122 List<Conversation> newValue = new ArrayList<>();
124 result.merge(hostname, newValue, (l1, l2) -> { l1.addAll(l2); return l1; });
126 if (hostnames.size() > 1) {
127 // Print notice of IP mapping to multiple hostnames (debugging)
128 System.err.println(String.format("%s: encountered an IP that maps to multiple (%d) hostnames",
129 TcpConversationUtils.class.getSimpleName(), hostnames.size()));
132 // If no hostname mapping, store conversation under the key that is the concatenation of the two IPs.
133 // In order to ensure consistency when mapping conversations, use lexicographic order to select which IP
135 String delimiter = "_";
136 // Note that the in case the comparison returns 0, the strings are equal, so it doesn't matter which of
137 // ipSrc and ipDst go first (also, this case should not occur in practice as it means that the device is
138 // communicating with itself!)
139 String key = ipSrc.compareTo(ipDst) <= 0 ? ipSrc + delimiter + ipDst : ipDst + delimiter + ipSrc;
140 List<Conversation> newValue = new ArrayList<>();
142 result.merge(key, newValue, (l1, l2) -> { l1.addAll(l2); return l1; });
148 public static Map<String, Integer> countPacketSequenceFrequencies(Collection<Conversation> conversations) {
149 Map<String, Integer> result = new HashMap<>();
150 for (Conversation conv : conversations) {
151 if (conv.getPackets().size() == 0) {
152 // Skip conversations with no payload packets.
155 StringBuilder sb = new StringBuilder();
156 for (PcapPacket pp : conv.getPackets()) {
157 sb.append(pp.length() + " ");
159 result.merge(sb.toString(), 1, (i1, i2) -> i1+i2);
165 * Given a {@link Collection} of {@link Conversation}s, builds a {@link Map} from {@link String} to {@link List}
166 * of {@link Conversation}s such that each key is the <em>concatenation of the packet lengths of all payload packets
167 * (i.e., the set of packets returned by {@link Conversation#getPackets()}) separated by a delimiter</em> of any
168 * {@link Conversation} pointed to by that key. In other words, what the {@link Conversation}s {@code cs} pointed to
169 * by the key {@code s} have in common is that they all contain exactly the same number of payload packets <em>and
170 * </em> these payload packets are identical across all {@code Conversation}s in {@code cs} in terms of packet
171 * length and packet order. For example, if the key is "152 440 550", this means that every individual
172 * {@code Conversation} in the list of {@code Conversation}s pointed to by that key contain exactly three payload
173 * packet of lengths 152, 440, and 550, and these three packets are ordered in the order prescribed by the key.
175 * @param conversations The collection of {@code Conversation}s to group by packet sequence.
176 * @param verbose If set to {@code true}, the grouping (and therefore the key) will also include SYN/SYNACK,
177 * FIN/FINACK, RST packets, and each payload-carrying packet will have an indication of the direction
178 * of the packet prepended.
179 * @return a {@link Map} from {@link String} to {@link List} of {@link Conversation}s such that each key is the
180 * <em>concatenation of the packet lengths of all payload packets (i.e., the set of packets returned by
181 * {@link Conversation#getPackets()}) separated by a delimiter</em> of any {@link Conversation} pointed to
184 public static Map<String, List<Conversation>> groupConversationsByPacketSequence(Collection<Conversation> conversations, boolean verbose) {
185 return conversations.stream().collect(Collectors.groupingBy(c -> toSequenceString(c, verbose)));
188 public static Map<String, List<Conversation>> groupConversationsByTlsApplicationDataPacketSequence(Collection<Conversation> conversations) {
189 return conversations.stream().collect(Collectors.groupingBy(
190 c -> c.getTlsApplicationDataPackets().stream().map(p -> Integer.toString(p.getOriginalLength())).
191 reduce("", (s1, s2) -> s1.length() == 0 ? s2 : s1 + " " + s2))
196 * Given a {@link Conversation}, counts the frequencies of each unique packet length seen as part of the
197 * {@code Conversation}.
198 * @param c The {@code Conversation} for which unique packet length frequencies are to be determined.
199 * @return A mapping from packet length to its frequency.
201 public static Map<Integer, Integer> countPacketLengthFrequencies(Conversation c) {
202 Map<Integer, Integer> result = new HashMap<>();
203 for (PcapPacket packet : c.getPackets()) {
204 result.merge(packet.length(), 1, (i1, i2) -> i1 + i2);
210 * Like {@link #countPacketLengthFrequencies(Conversation)}, but counts packet length frequencies for a collection
211 * of {@code Conversation}s, i.e., the frequency of a packet length becomes the total number of packets with that
212 * length across <em>all</em> {@code Conversation}s in {@code conversations}.
213 * @param conversations The collection of {@code Conversation}s for which packet length frequencies are to be
215 * @return A mapping from packet length to its frequency.
217 public static Map<Integer, Integer> countPacketLengthFrequencies(Collection<Conversation> conversations) {
218 Map<Integer, Integer> result = new HashMap<>();
219 for (Conversation c : conversations) {
220 Map<Integer, Integer> intermediateResult = countPacketLengthFrequencies(c);
221 for (Map.Entry<Integer, Integer> entry : intermediateResult.entrySet()) {
222 result.merge(entry.getKey(), entry.getValue(), (i1, i2) -> i1 + i2);
228 public static Map<String, Integer> countPacketPairFrequencies(Collection<PcapPacketPair> pairs) {
229 Map<String, Integer> result = new HashMap<>();
230 for (PcapPacketPair ppp : pairs) {
231 result.merge(ppp.toString(), 1, (i1, i2) -> i1 + i2);
236 public static Map<String, Map<String, Integer>> countPacketPairFrequenciesByHostname(Collection<Conversation> tcpConversations, DnsMap ipHostnameMappings) {
237 Map<String, List<Conversation>> convsByHostname = groupConversationsByHostname(tcpConversations, ipHostnameMappings);
238 HashMap<String, Map<String, Integer>> result = new HashMap<>();
239 for (Map.Entry<String, List<Conversation>> entry : convsByHostname.entrySet()) {
240 // Merge all packet pairs exchanged during the course of all conversations with hostname into one list
241 List<PcapPacketPair> allPairsExchangedWithHostname = new ArrayList<>();
242 entry.getValue().forEach(conversation -> allPairsExchangedWithHostname.addAll(extractPacketPairs(conversation)));
243 // Then count the frequencies of packet pairs exchanged with the hostname, irrespective of individual
245 result.put(entry.getKey(), countPacketPairFrequencies(allPairsExchangedWithHostname));
251 * Given a {@link Conversation}, extract its packet length sequence.
252 * @param c The {@link Conversation} from which a packet length sequence is to be extracted.
253 * @return An {@code Integer[]} that holds the packet lengths of all payload-carrying packets in {@code c}. The
254 * packet lengths in the returned array are ordered by packet timestamp.
256 public static Integer[] getPacketLengthSequence(Conversation c) {
257 return getPacketLengthSequence(c.getPackets());
262 * Given a {@link Conversation}, extract its packet length sequence, but only include packet lengths of those
263 * packets that carry TLS Application Data.
264 * @param c The {@link Conversation} from which a TLS Application Data packet length sequence is to be extracted.
265 * @return An {@code Integer[]} that holds the packet lengths of all packets in {@code c} that carry TLS Application
266 * Data. The packet lengths in the returned array are ordered by packet timestamp.
268 public static Integer[] getPacketLengthSequenceTlsAppDataOnly(Conversation c) {
270 throw new IllegalArgumentException("Provided " + c.getClass().getSimpleName() + " was not a TLS session");
272 return getPacketLengthSequence(c.getTlsApplicationDataPackets());
276 * Given a list of packets, extract the packet lengths and wrap them in an array such that the packet lengths in the
277 * resulting array appear in the same order as their corresponding packets in the input list.
278 * @param packets The list of packets for which the packet lengths are to be extracted.
279 * @return An array containing the packet lengths in the same order as their corresponding packets in the input list.
281 private static Integer[] getPacketLengthSequence(List<PcapPacket> packets) {
282 return packets.stream().map(pkt -> pkt.getOriginalLength()).toArray(Integer[]::new);
286 * Builds a string representation of the sequence of packets exchanged as part of {@code c}.
287 * @param c The {@link Conversation} for which a string representation of the packet sequence is to be constructed.
288 * @param verbose {@code true} if set to true, the returned sequence string will also include SYN/SYNACK,
289 * FIN/FINACK, RST packets, as well as an indication of the direction of payload-carrying packets.
290 * @return a string representation of the sequence of packets exchanged as part of {@code c}.
292 private static String toSequenceString(Conversation c, boolean verbose) {
293 // Payload-parrying packets are always included, but only prepend direction if verbose output is chosen.
294 Stream<String> s = c.getPackets().stream().map(p -> verbose ? c.getDirection(p).toCompactString() + p.getOriginalLength() : Integer.toString(p.getOriginalLength()));
296 // In the verbose case, we also print SYN, FIN and RST packets.
297 // Convert the SYN packets to a string representation and prepend them in front of the payload packets.
298 s = Stream.concat(c.getSynPackets().stream().map(p -> isSyn(p) && isAck(p) ? "SYNACK" : "SYN"), s);
299 // Convert the FIN packets to a string representation and append them after the payload packets.
300 s = Stream.concat(s, c.getFinAckPairs().stream().map(f -> f.isAcknowledged() ? "FINACK" : "FIN"));
301 // Convert the RST packets to a string representation and append at the end.
302 s = Stream.concat(s, c.getRstPackets().stream().map(r -> "RST"));
305 * Note: the collector internally uses a StringBuilder, which is more efficient than simply doing string
306 * concatenation as in the following example:
307 * s.reduce("", (s1, s2) -> s1.length() == 0 ? s2 : s1 + " " + s2);
308 * (above code is O(N^2) where N is the number of characters)
310 return s.collect(Collectors.joining(" "));
314 * Appends a space to {@code sb} <em>iff</em> {@code sb} already contains some content.
315 * @param sb A {@link StringBuilder} that should have a space appended <em>iff</em> it is not empty.
317 private static void appendSpaceIfNotEmpty(StringBuilder sb) {
318 if (sb.length() != 0) {