diff --git a/nflux-ebpf/src/main.rs b/nflux-ebpf/src/main.rs
index 23bfb6d..fad1d8c 100644
--- a/nflux-ebpf/src/main.rs
+++ b/nflux-ebpf/src/main.rs
@@ -3,7 +3,7 @@
 #![allow(nonstandard_style, dead_code)]
 
 use aya_ebpf::maps::lpm_trie::Key;
-use aya_ebpf::maps::{Array, LpmTrie};
+use aya_ebpf::maps::{Array, LpmTrie, LruHashMap};
 use aya_ebpf::{
     bindings::xdp_action,
     macros::{map, xdp},
@@ -38,6 +38,9 @@ static ICMP_RULE: Array<u32> = Array::with_max_entries(1, 0);
 #[map]
 static EGRESS_EVENT: PerfEventArray<EgressEvent> = PerfEventArray::new(0);
 
+#[map]
+static ACTIVE_CONNECTIONS: LruHashMap<u32, u32> = LruHashMap::with_max_entries(1024, 0);
+
 #[xdp]
 pub fn nflux(ctx: XdpContext) -> u32 {
     match start_nflux(ctx) {
@@ -78,20 +81,27 @@ fn log_new_connection(ctx: XdpContext, src_addr: u32, dst_port: u16, protocol: u
 fn try_tc_egress(ctx: TcContext) -> Result<i32, ()> {
     let ethhdr: EthHdr = ctx.load(0).map_err(|_| ())?;
     match ethhdr.ether_type {
-        EtherType::Ipv4 => {}
+        EtherType::Ipv4 => unsafe {
+            let ipv4hdr: Ipv4Hdr = ctx.load(EthHdr::LEN).map_err(|_| ())?;
+            let destination = u32::from_be(ipv4hdr.dst_addr);
+
+            // Check if this destination is already active
+            if ACTIVE_CONNECTIONS.get(&destination).is_none() {
+                // Log only new connections
+                let event = EgressEvent { dst_ip: destination };
+                EGRESS_EVENT.output(&ctx, &event, 0);
+
+                // Mark connection as active
+                ACTIVE_CONNECTIONS.insert(&destination, &1, 0).map_err(|_| ())?;
+            }
+        }
         _ => return Ok(TC_ACT_PIPE),
     }
 
-    let ipv4hdr: Ipv4Hdr = ctx.load(EthHdr::LEN).map_err(|_| ())?;
-    let destination = u32::from_be(ipv4hdr.dst_addr);
-
-    let event = EgressEvent { dst_ip: destination };
-
-    EGRESS_EVENT.output(&ctx, &event, 0);
-
     Ok(TC_ACT_PIPE)
 }
 
+
 fn start_nflux(ctx: XdpContext) -> Result<u32, ()> {
     let ethhdr: *const EthHdr = unsafe { ptr_at(&ctx, 0)? };
 
diff --git a/nflux/src/main.rs b/nflux/src/main.rs
index 3368bc3..d879a88 100644
--- a/nflux/src/main.rs
+++ b/nflux/src/main.rs
@@ -15,7 +15,7 @@ use bytes::BytesMut;
 use config::{Action, Nflux, Protocol, IpRules};
 use core::set_mem_limit;
 use logger::setup_logger;
-use nflux_common::{convert_protocol, ConnectionEvent, IpRule, LpmKeyIpv4, LpmKeyIpv6};
+use nflux_common::{convert_protocol, ConnectionEvent, EgressEvent, IpRule, LpmKeyIpv4, LpmKeyIpv6};
 use std::collections::HashMap;
 use std::net::{Ipv4Addr, Ipv6Addr};
 use std::ptr;
@@ -82,22 +82,58 @@ async fn main() -> anyhow::Result<()> {
     )?;
 
     let mut egress_events = AsyncPerfEventArray::try_from(
-        bpf.take_map("EGRESS_EVENTS")
-            .context("Failed to find EGRESS_EVENTS map")?,
+        bpf.take_map("EGRESS_EVENT")
+            .context("Failed to find EGRESS_EVENT map")?,
     )?;
 
     let cpus = online_cpus().map_err(|(_, error)| error)?;
 
     for cpu_id in cpus {
-        let buf = events.open(cpu_id, None)?;
-        task::spawn(process_events(buf, cpu_id));
+        // Spawn task for connection events
+        {
+            let buf = events.open(cpu_id, None)?;
+            task::spawn(process_events(buf, cpu_id));
+        }
+
+        // Spawn task for egress events
+        {
+            let buf = egress_events.open(cpu_id, None)?;
+            task::spawn(process_egress_events(buf, cpu_id));
+        }
     }
 
+
     // Wait for shutdown signal
     wait_for_shutdown().await?;
     Ok(())
 }
 
+async fn process_egress_events(
+    mut buf: AsyncPerfEventArrayBuffer<MapData>,
+    cpu_id: u32,
+) -> Result<(), PerfBufferError> {
+    let mut buffers = vec![BytesMut::with_capacity(1024); 10];
+
+    loop {
+        // Wait for events
+        let events = buf.read_events(&mut buffers).await?;
+
+        // Process each event in the buffer
+        for i in 0..events.read {
+            let buf = &buffers[i];
+            match parse_egress_event(buf) {
+                Ok(event) => {
+                    info!(
+                        "direction=outgoing ip={}",
+                        Ipv4Addr::from(event.dst_ip)
+                    );
+                }
+                Err(e) => error!("Failed to parse egress event on CPU {}: {}", cpu_id, e),
+            }
+        }
+    }
+}
+
 async fn process_events(
     mut buf: AsyncPerfEventArrayBuffer<MapData>,
     cpu_id: u32,
@@ -127,6 +163,18 @@ async fn process_events(
     }
 }
 
+fn parse_egress_event(buf: &BytesMut) -> anyhow::Result<EgressEvent> {
+    if buf.len() >= std::mem::size_of::<EgressEvent>() {
+        let ptr = buf.as_ptr() as *const EgressEvent;
+        let event = unsafe { ptr::read_unaligned(ptr) };
+        Ok(event)
+    } else {
+        Err(anyhow::anyhow!(
+            "Buffer size is too small for EgressEvent"
+        ))
+    }
+}
+
 fn parse_connection_event(buf: &BytesMut) -> anyhow::Result<ConnectionEvent> {
     if buf.len() >= std::mem::size_of::<ConnectionEvent>() {
         let ptr = buf.as_ptr() as *const ConnectionEvent;